Experimental Characterisation and Numerical Simulation of ﬁbre Laser Welding of AA 2024-T3 and Ti-6Al-4V

Home , 2024 aluminium alloy

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/309813991

Experimental characterisation and numerical simulation of ﬁbre laser welding of AA 2024-T3 and Ti-6Al-4V

Thesis · October 2016 DOI: 10.13140/RG.2.2.26191.07848

CITATIONS READS 2 569

1 author:

Joseph Ahn Samsung

17 PUBLICATIONS 404 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Structural integrity of high power fibre laser and electron beam welded lightweight alloy aircraft structures View project

All content following this page was uploaded by Joseph Ahn on 10 November 2016.

The user has requested enhancement of the downloaded file.

EXPERIMENTAL CHARACTERISATION AND NUMERICAL SIMULATION OF FIBRE LASER WELDING OF AA 2024-T3 AND Ti-6Al-4V

A Thesis submitted for the Degree of Doctor of Philosophy of

Imperial College London

and

Diploma of Imperial College

Joseph Ahn

June 2016

Department of Mechanical Engineering

Imperial College London

SW7 2AZ ABSTRACT

The aircraft industry has long recognised the importance of climate protection and the benefits of reducing weight for the production of cost effective and fuel efficient aircraft structures. Fibre laser welding provides advantages over conventional riveting, mainly in terms of weight reduction and time saving. However, significant changes in microstructure, metallurgical state and associated mechanical properties occur in welded joints. Such changes can result in residual stresses, distortions and defects formation in the welded structure, thus significantly influencing the performance and service life. In order to maintain structural integrity of welded structures, the relationship between welding process and performance of the structure needs to be fully assessed. In this thesis, comprehensive relationships between materials, welding process, microstructure and mechanical properties of welded joints were established.

Welding parameters including power density, laser power, welding speed, focal position, filler metal feed rate and shielding gas composition were optimised to produce high quality full penetration welds. Solidification cracking was found to be a critical issue in AA 2024-T3 when welding without filler metal. The addition of filler metal reduced its crack sensitivity but it was also necessary to provide the optimum feed rate to avoid welding defects and keyhole instability. Sufficiently high laser power and low welding speed were required for full penetration and also to minimise welding defects. Both argon and helium shielding gases were found to be effective since only weakly ionised laser induced vapour plume was formed rather than strongly ionised plasma.

Softening in AA204-T3 deteriorated the plastic straining capacity of the weld due to confined plasticity development within the weld. A poor weld quality resulted in a mixed mode of brittle and ductile failure and contained micro porosities and hot cracks, whereas, a good weld quality led to a ductile mode with significantly less welding defects. In the case of Ti-6Al-4V, the strength was the greatest in the weld as a result of martensitic microstructure formed during fast cooling rates. Local plastic deformation was the lowest in the weld and therefore, failed in the parent material but at the cost of reduced ductility relative to the unwelded parent tensile specimens.

The residual stresses and distortions due to time dependent and localised heating imposed during fibre laser welding were numerically simulated with thermal and mechanical boundary conditions integrated in the finite element models including post weld heat treatment, mechanical stress relieving treatment and various clamping arrangements. Mechanical boundary conditions had relatively small influence on residual stresses in thin sheets of butt

i welded specimens, whereas, greater restraints led to higher residual stresses and lower restraints led to lower residual stresses in T-joint specimens.

Non-isothermal diffusional and diffusionless phase transformations in Ti-6Al-4V were modelled and their influence on residual stresses and distortions was examined. Phase transformations only had a small influence on the magnitude and distribution of residual stresses and distortions because the level of internal stresses due to phase transformation remained low unlike other materials which exhibit greater differences in the specific volumes between phases.

Post weld heat treatment (PWHT) induced diffusional phase transformations via decomposition of martensite into α. It also decreased the magnitude of y stresses to the yield strength of Ti-6Al-4V at the treatment temperature by releasing the locked-in stresses. Mechanical stress relieving was also studied for reducing residual stresses and distortions, by means of plastic deformation applied during as well as after welding. When the load reached more than 50% of its yield strength, the stresses became compressive.

Residual stresses were experimentally measured using X-ray and neutron diffraction techniques They were found to be dependent on the crystallographic hkl plane due to the presence of microscopic stresses. In the case of Ti-6Al-4V, the reflections were weak and only few times larger than the background due its highly incoherent cross-section. In addition, texture in Ti-6Al-4V weld also contributed to lower intensity counts observed during measurements. As a result, only certain peaks were detected in certain orientations. The Y residual stresses in the welding direction were very high but not as high as the yield strength of the material at room temperature for both AA 2024-T3 and Ti-6Al-4V. They were largely tensile in nature only within the weld and tended to be weakly compressive in the rest of the specimen. Comparative analyses between experimental and numerical results showed good agreements, proving the robustness of the finite element models.

ii ACKNOWLEDGMENTS

I would like to express my gratitude to my academic supervisors Professor John Dear and Dr Catrin Davies for their support, guidance and encouragement. I owe my thanks also to Professor Li Chen and Mr Enguang He for their technical assistance, extended cooperation and suggestions.

I wish to also thank my past and present colleagues and members of staff at Imperial College London for their support, and instrument scientists from AVIC BAMTRI, ILL, ISIS and HZB for their willingness to share their knowledge and extending their assistance in performing experimental tasks throughout this project.

Personally, I would like to sincerely thank to my father, mother and sister for their support and encouragement. Special thanks to my wife, for her support and encouragement during this time.

iii DECLARATION OF ORIGINALITY

The research conducted within this document has been undertaken by the author except where explicitly stated. Where the products of other works have been used, they have been appropriately referenced both in the text and the accompanying bibliography.

The copyright of this thesis rests with the author and is made available under a Creative Commons Attribution Non-Commercial No Derivatives licence. Researchers are free to copy, distribute or transmit the thesis on the condition that they attribute it, that they do not use it for commercial purposes and that they do not alter, transform or build upon it. For any reuse or redistribution, researchers must make clear to others the licence terms of this work.

iv TABLE OF CONTENTS

ABSTRACT ...... i

ACKNOWLEDGMENTS ...... iii

DECLARATION OF ORIGINALITY ...... iv

TABLE OF CONTENTS ...... v

LIST OF FIGURES ...... ix

LIST OF TABLES ...... xxiii

NOMENCLATURE ...... xxiv

1 INTRODUCTION ...... 1

1.1 Aims and Objectives ...... 3

1.2 Structure of the Thesis ...... 4

2 LITERATURE REVIEW ...... 5

2.1 Introduction ...... 5

2.2 Structure and Properties of AA 2024-T3 ...... 5

2.2.1 The Use of AA 2024-T3 in Aircraft Applications ...... 10

2.3 Structure and Properties of Ti-6Al-4V ...... 12

2.3.1 The Use of Ti-6Al-4V in Aircraft Applications ...... 14

2.4 Laser Beam Welding Principles ...... 17

2.4.1 Conduction Mode and Keyhole Mode ...... 18

2.4.2 Continuous Wave and Pulse Wave Operation ...... 20

2.5 Laser Beam Welding Processes ...... 20

2.5.1 Friction Stir Welding Process ...... 24

2.6 Research Progress on Laser Welding of AA 2024-T3 ...... 24

2.6.1 Fibre Laser Welding of AA 2024-T3 ...... 25

2.7 Research Progress on Laser Welding of Ti-6Al-4V ...... 26

2.7.1 Fibre Laser Welding of Ti-6Al-4V ...... 28

2.8 Conclusions ...... 29

v 3 EFFECT OF WELDING ON MICROSTRUCTURE OF AA 2024-T3 AND Ti-6Al-4V .... 31

3.1 Introduction ...... 31

3.2 Materials and Experimental Procedures ...... 34

3.2.1 Materials ...... 34

3.2.2 Fibre Laser Beam Welding Process ...... 35

3.2.3 Metallographic Specimen Preparation ...... 38

3.2.4 Experimental Procedures ...... 39

3.2.5 Welding Quality Acceptance Criteria ...... 41

3.3 Results and Discussion on AA 2024-T3 ...... 42

3.3.1 Effect of Power Density ...... 47

3.3.2 Effect of Laser Power ...... 52

3.3.3 Effect of Welding Speed ...... 59

3.3.4 Effect of Focal Position ...... 67

3.3.5 Effect of Filler Metal Feed Rate ...... 76

3.3.6 Effect of Shielding Gas ...... 81

3.3.7 Fillet Welding of AA 2024-T3 T-joints ...... 87

3.4 Results and Discussion on Ti-6Al-4V ...... 89

3.4.1 Effect of Laser Power ...... 92

3.4.2 Effect of Welding Speed ...... 100

3.4.3 Effect of Focal Position ...... 108

3.4.4 Fillet Welding of Ti-6Al-4V T-joints ...... 110

3.5 Conclusions ...... 112

3.5.1 AA 2024-T3 ...... 112

3.5.2 Ti-6Al-4V ...... 113

4 EFFECT OF WELDING ON MECHANICAL PROPERTIES OF AA 2024-T3 AND Ti-6Al- 4V ...... 114

4.1 Introduction ...... 114

4.1.1 Digital Image Correlation ...... 115

4.2 Materials and Experimental Procedures ...... 116

vi 4.3 Results and Discussion on AA 2024-T3 ...... 120

4.3.1 Micro-indentation Hardness ...... 120

4.3.2 Global and Local Tensile Properties ...... 123

4.4 Results and Discussion on Ti-6Al-4V ...... 139

4.4.1 Micro-indentation Hardness ...... 139

4.4.2 Global and Local Tensile Properties ...... 141

4.4.3 Preliminary Physical Simulation of Ti-6Al-4V Weld Microstructure ...... 150

4.5 Conclusions ...... 154

4.5.1 AA 2024-T3 ...... 154

4.5.2 Ti-6Al-4V ...... 155

5 EXPERIMENTAL MEASUREMENT OF WELDING RESIDUAL STRESSES AND DISTORTIONS IN AA 2024-T3 AND Ti-6Al-4V ...... 156

5.1 Introduction ...... 156

5.2 Experimental Measurement of Welding Distortions ...... 159

5.3 Experimental Measurement of Welding Residual Stresses ...... 159

5.3.1 X-ray Diffraction ...... 162

5.3.2 Neutron Diffraction ...... 163

5.4 Results and Discussion on AA 2024-T3 ...... 170

5.5 Results and Discussion on Ti-6Al-4V ...... 181

5.6 Conclusions ...... 186

6 NUMERICAL SIMULATION OF WELDING RESIDUAL STRESSES AND DISTORTIONS IN AA 2024-T3 AND Ti-6Al-4V ...... 188

6.1 Introduction ...... 188

6.2 Materials and Welding Procedures ...... 192

6.3 Finite Element Model Geometry and Mesh ...... 196

6.4 Thermal Analysis ...... 198

6.5 Mechanical Analysis ...... 203

6.6 Results and Discussion on AA 2024-T3 ...... 207

6.6.1 Simulated Thermal Histories and Weld Seam Geometry ...... 207

6.6.2 Simulated Residual Stresses and Out of Plane Displacements ...... 211

vii 6.7 Results and Discussion on Ti-6Al-4V ...... 228

6.7.1 Simulated Thermal Histories and Weld Seam Geometry ...... 228

6.7.2 Simulated Residual Stresses and Out of Plane Displacements ...... 231

6.7.3 Modelling Solid State Phase Transformations in Ti-6Al-4V ...... 233

6.7.4 Mechanical Stress Relieving (MSR) Treatment...... 246

6.8 Conclusions ...... 249

7 CONCLUSIONS AND FUTURE WORK ...... 251

7.1 Conclusions ...... 251

7.2 Future Work ...... 254

REFERENCES ...... 256

LIST OF PUBLICATIONS ...... 295

APPENDICES ...... 296

A. Fibre Laser Welding Parameters Utilised ...... 296

B. Miscellaneous Results ...... 300

C. User Defined Subroutines used in Finite Element Modelling ...... 313

viii LIST OF FIGURES Figure 1 Reflectivity of aluminium and other metals as a function of laser wavelength for normally incident light on a flat surface at room temperature [26,27] ...... 7

Figure 2 Structural weight ratio of different materials in commercial aircraft airframe and engine [74]...... 16

Figure 3 Intensity distributions for various beam modes ...... 35

Figure 4 a) Experimental setup and b) a cross-sectional view of the fixture used for automated fibre laser welding of AA 2024-T3 and Ti-6Al-4V thin sheets ...... 36

Figure 5 Conformity of weld seam geometry represented by a) a good weld surface quality and b) a poor weld surface quality...... 40

Figure 6 Operating welding process parameters window of laser power and welding speed for AA 2024-T3 ...... 44

Figure 7 Microstructures of fibre laser welded 3 mm thick AA 2024-T3 in a) the weld at 25x, b) the base metal at 100x, c) the HAZ/FZ boundary at 100x, d) the heat affected zone at 500x, e) equiaxed dendrites in the fusion zone at 200x, and f) columnar dendrites in the fusion zone at 200x magnifications...... 45

Figure 8 Influence of power density on a) top and bottom weld widths and b) weld width ratio, undercut and reinforcement ...... 49

Figure 9 Transverse sections and weld top bead profiles produced with different power densities ...... 49

Figure 10 Microstructure of the AA 2024-T3 fusion zone with equiaxed dendrite structure at the weld centreline and columnar dendrites on each side at 50x, and detailed images of solidification cracks at 500x magnifications for specimens with different power densities of a) 2.05, b) 2.76 and c) 3.47 MW/cm2 ...... 50

Figure 11 Weight percentage (%) of chemical elements in base metal and welds of different power densities ...... 52

Figure 12 a) Relationship between weld width and laser power at a welding speed of 2.0 m/min with no defocus and with argon shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement ...... 53

Figure 13 Transverse sections and weld top bead profiles produced with different laser powers at a welding speed of 2.0 m/min with no defocus and argon shielding gas ...... 53

ix Figure 14 a) Relationship between weld width and laser power at a welding speed of 3.0 m/min with no defocus and with argon shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement ...... 55

Figure 15 Transverse sections and weld top bead profiles produced with different laser powers at a welding speed of 3.0 m/min with no defocus and argon shielding gas ...... 55

Figure 16 a) Relationship between weld width and laser power at a welding speed of 2.0 m/min with +4 mm defocus and with argon shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement ...... 58

Figure 17 Transverse sections and weld top bead profiles produced with different laser powers at a welding speed of 2.0 m/min with +4 mm defocus and argon shielding gas ...... 58

Figure 18 a) Relationship between weld width and welding speed at a laser power of 1.9 kW with +4 mm defocus and with argon shielding gas, and b) the resultant underfill and reinforcement defects ...... 61

Figure 19 Transverse sections and weld top bead profiles produced with different welding speeds at a laser power of 1.9 kW with +4 mm defocus and argon shielding gas ...... 61

Figure 20 a) Relationship between weld width and welding speed at a laser power of 2.9 kW

with +4 mm defocus and with argon shielding gas, and b) the resultant Rw, undercut, underfill and reinforcement defects ...... 63

Figure 21 Transverse sections and weld top bead profiles produced with different welding speeds at a laser power of 2.9 kW with +4 mm defocus and argon shielding gas ...... 63

Figure 22 a) Relationship between weld width and welding speed at a laser power of 4.9 kW

with +4 mm defocus and with argon shielding gas, and b) the resultant Rw, undercut, underfill and reinforcement defects ...... 66

Figure 23 Transverse sections and weld top bead profiles produced with different welding speeds at a laser power of 4.9 kW with +4 mm defocus and argon shielding gas ...... 66

Figure 24 a) Relationship between weld width and focal position at a laser power of 3.9 kW, a welding speed of 2.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement ...... 70

Figure 25 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 3.9 kW, a welding speed of 2.0 m/min and with argon shielding gas ...... 70

x Figure 26 a) Relationship between weld width and focal position at a laser power of 4.9 kW, a welding speed of 3.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement ...... 72

Figure 27 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 4.9 kW, a welding speed of 3.0 m/min and with argon shielding gas ...... 72

Figure 28 a) Relationship between weld width and focal position at a laser power of 4.9 kW, a welding speed of 3.0 m/min, a filler metal feed rate of 5.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement ...... 75

Figure 29 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 4.9 kW, a welding speed of 3.0 m/min, a filler metal feed rate of 5.0 m/min and Ar ...... 75

Figure 30 a) Relationship between weld width, magnesium and silicon content and filler metal feed rate at a laser power of 4.9 kW, a welding speed of 3.0 m/min, +4 mm defocus and with helium shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement ...... 78

Figure 31 Transverse sections of welds and weld top bead profiles produced with different filler metal feed rate at a laser power of 4.9 kW, a welding speed of 3.0 m/min, +4 mm defocus and with helium shielding gas ...... 79

Figure 32 a) Weight percentage (%) of main alloying elements in the weld as a function of filler metal feed rate obtained using energy dispersive X-ray spectroscopy (EDX) and b) aluminium crack sensitivity curves showing the effects of different alloy additions [246,271]...... 81

Figure 33 a) Relationship between weld width and welding speed at a laser power of 4.9 kW, no defocus and with either argon or helium shielding gas, and b) the resultant Rw, undercut, underfill and reinforcement defects ...... 84

Figure 34 Transverse sections of welds and weld top bead profiles produced with different welding speeds at a laser power of 4.9 kW, no defocus and with either argon or helium shielding gas ...... 85

Figure 35 a) Relationship between weld width and focal position at a laser power of 4.9 kW, a welding speed of 3.0 m/min and with either argon or helium shielding gas, and b) the resultant weld width ratio, undercut and reinforcement...... 86

xi Figure 36 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 4.9 kW, at a welding speed of 3.0 m/min and with either argon or helium shielding gas ...... 87

Figure 37 Microstructure of two pass fillet welded AA 2024-T3 T-joints a) showing weld profile without defect b) weld profile extracted from another position showing spatter, root porosity and underfill, c) in the weld containing micro porosity at 100x, d) in the weld containing root macro porosity at 100x, e) in the fusion zone at 500x and f) in the heat affected zone at 500x magnifications, etched using Kroll’s reagent ...... 88

Figure 38 Microstructure of fibre laser welded Ti-6Al-4V in a) the base metal, b) the heat affected zone and c) the fusion zone at 200x and 500x magnifications etched using Kroll’s reagent ...... 91

Figure 39 a) Energy-dispersive X-ray spectroscopy (EDX) spectrum of the fusion zone and b) the detected weight percentage (%) of chemical elements in the parent material, the heat affected zone and the fusion zone ...... 92

Figure 40 a) Relationship between weld width and laser power at a welding speed of 2.1 m/min with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 94

Figure 41 Transverse sections of welds produced with different laser powers at a welding speed of 2.1 m/min with +4 mm defocus and argon shielding gas...... 94

Figure 42 a) Relationship between weld width and laser power at a welding speed of 3.9 m/min with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 97

Figure 43 Transverse sections of welds produced with different laser powers at a welding speed of 3.9 m/min with +4 mm defocus and argon shielding gas...... 97

Figure 44 a) Relationship between weld width and laser power at a welding speed of 6.0 m/min with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 99

Figure 45 Transverse sections of welds produced with different laser powers at a welding speed of 6.0 m/min with +4 mm defocus and argon shielding gas...... 99

Figure 46 a) Relationship between weld width and welding speed at a laser power of 2.0 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 101

xii Figure 47 Transverse sections of welds produced with different welding speeds at a laser power of 2.0 kW with +4 mm defocus and argon shielding gas ...... 101

Figure 48 a) Relationship between weld width and welding speed at a laser power of 2.5 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 103

Figure 49 Transverse sections of welds produced with different welding speeds at a laser power of 2.5 kW with +4 mm defocus and argon shielding gas ...... 103

Figure 50 a) Relationship between weld width and welding speed at a laser power of 3.0 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 105

Figure 51 Transverse sections of welds produced with different welding speeds at a laser power of 3.0 kW with +4 mm defocus and argon shielding gas ...... 105

Figure 52 a) Relationship between weld width and welding speed at a laser power of 3.5 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement ...... 107

Figure 53 Transverse sections of welds produced with different welding speeds at a laser power of 3.5 kW with +4 mm defocus and argon shielding gas ...... 107

Figure 54 a) Relationship between weld width and focal position at a laser power of 2.5 kW, a welding speed of 3.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement ...... 109

Figure 55 Transverse sections of welds produced with different focal positions at a laser power of 2.5 kW, a welding speed of 3.0 m/min and with argon shielding gas ...... 110

Figure 56 Microstructure of two pass fillet welded Ti-6Al-4V a) overall, b) in the weld at 50x, c) in the top heat affected zone at 100x, d) in the fusion zone at 200x, e) in the bottom heat affected zone at 200x and f) the porosity defects in the fusion boundary at 500x magnifications, etched using Kroll’s reagent ...... 111

Figure 57 An additional clamping system adopted to perform the test on the adhesion of the stiffener to the skin ...... 119

Figure 58 Effect of welding speed on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds ...... 122

Figure 59 Effect of wire feed rate on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds ...... 122

xiii Figure 60 Effect of focal distance on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds ...... 122

Figure 61 Joint efficiencies of welded macro-tensile specimens for the corresponding welding parameters listed in Table 9 ...... 125

Figure 62 Full field y strain distributions in loading direction for fibre laser welded AA 2024-T3 processed under four different welding conditions, showing the development of strain localisation relative to the time to fracture ...... 126

Figure 63 Global mechanical responses obtained by standard tensile tests for welded specimens of four different welding conditions showing a) engineering stress and strain curves, b) true stress and strain curves, c) elastic modulus and elongation to failure and d) yield strength and ultimate tensile strength ...... 128

Figure 64 Local mechanical responses in FZ and HAZ constructed from full field DIC tensile tests compared to overal responses for four different welding conditions ...... 129

Figure 65 SEM fracture morphology of the DIC tensile test specimens for the four different welding conditions at 500x magnification showing different modes of failure and presence of welding defects ...... 131

Figure 66 DIC computed global tensile response of micro-tensile specimens a) compared to macro-tensile specimens and for different b) laser powers, c) welding speeds and d) focal positions ...... 133

Figure 67 DIC Total strain field of fibre laser welded AA 2024-T3 micro-tensile specimens at a welding speed of 2.7 m/min just before fracture at 50x magnification, local and global constitutive data determined for the various weld regions, and development of strain distribution across the weld during load to final fracture at different percentage of the fracture time ...... 135

Figure 68 Processed and raw DIC image data of fibre laser welded AA 2024-T3 longitudinal T-joint tensile specimens at 50%, 75%, 99% and 100% of the time to failure ...... 136

Figure 69 Processed and raw DIC image data of fibre laser welded AA 2024-T3 transverse T- joint tensile specimens at 50%, 75%, 99% and 100% of the time to failure ...... 137

Figure 70 Processed and raw DIC image data of fibre laser welded AA2024-T3 T-joint stiffener tension specimen at 50%, 75%, 99% and 100% of the time to failure ...... 138

Figure 71 Global tensile properties of a) transverse and longitudinal fillet welded T-joint, butt welded and parent tensile specimens and b) the force and displacement results of stiffener tension specimens extracted from three different locations of the workpiece 138

xiv Figure 72 Micro-indentation hardness distributions of fibre laser welded Ti-6Al-4V autogenous welds ...... 141

Figure 73 Full field y strain distributions in loading direction for fibre laser welded Ti-6Al-4V showing the development of strain localisations at different fraction of the time to fracture ...... 142

Figure 74 a) Local stress and strain behaviour in Ti-6Al-4V FZ, HAZ and BM, and global constitutive behaviour of the welded specimen, b) strain hardening parameters K and n of FZ and HAZ, c) predicted local plastic constitutive behaviour of FZ and HAZ...... 143

Figure 75 Global tensile properties of Ti-6Al-4V fibre laser welds processed under different laser power, welding speed and focal position ...... 146

Figure 76 Processed and raw DIC image data of fibre laser welded Ti-6Al-4V micro-tensile specimens as a function of laser power just before fracture and the equivalent chemically etched specimens without speckle patterns showing the location of fracture in the base metal ...... 147

Figure 77 Processed and raw DIC image data of fibre laser welded Ti-6Al-4V a) transverse and b) longitudinal T-joint tensile specimens at 50%, 75%, 99% and 100% of the time to failure ...... 149

Figure 78 Processed and raw DIC image data of fibre laser welded Ti-6Al-4V T-joint stiffener tension specimen at 50%, 75%, 99% and 100% of the time to failure ...... 149

Figure 79 Global tensile properties of a) transverse and longitudinal T-joint tensile specimens compared to butt welded and parent tensile specimens and b) stiffener tension specimens extracted from three different locations of the workpiece ...... 150

Figure 80 Thermo-mechanical simulation using the Gleeble showing a) programmed and measured thermal cycle for differernt cooling rates and b) stress and strain behaviour of simulated Ti-6Al-4V specimens subjected to tensile static loading under vacuum and at room termpature...... 151

Figure 81 Microstructures of unwelded Ti-6Al-4V after heating to above the β transus followed by cooling at different rates from 20 to 400C°s-1 to cause differernt transformation kinetics on cooling, compared to FZ microstructure obtained from an actual weld cross-section, at 100x magnification ...... 153

Figure 82 Evolution of temperature and stress in the welding direction during welding (figure taken from [335]) ...... 156

xv Figure 83 Experimental setup for measuring the out-of-plane displacement after welding using coordinate measuring machine ...... 159

Figure 84 X-ray diffractometer equipped with a copper X-ray source ...... 162

Figure 85 Diffraction spectrum of fibre laser welded Ti-6Al-4V from reactor based diffractometer E3 ...... 164

Figure 86 Diffraction spectrum of fibre laser welded AA 2024-T3 from pulsed spallation source ENGIN-X fitted with Rietveld refinements ...... 165

Figure 87 Residual stress measurement positions for butt welded AA 2024-T3 and Ti-6Al-4V sheets ...... 167

Figure 88 Residual stress measurement positions for T-joint fillet welded AA 2024-T3 thin sheets with a single stiffener and triple stiffeners ...... 167

Figure 89 Dimensions of electrical discharge machined (EDM) stress free reference combs ...... 168

Figure 90 Butt welded AA 2024-T3 specimen mounted on the sample table of the ENGIN-X neutron diffractor at ISIS and Ti-6Al-4V specimen mounted on the sample table of the E3 neutron diffraction meter at HZB ...... 168

Figure 91 T-joint fillet welded AA 2024-T3 specimens mounted on the hexapod platform of the SALSA neutron diffractometer at ILL ...... 169

Figure 92 In plane residual stress distributions on the top surface of butt welded AA 2024-T3 sheets in x and y directions measured using X-ray diffraction technique and from numerical simulations under three different sets of welding parameters a) W47, b) W99 and c) W102 (AA 2024-T3, butt joint) (AA 2024-T3, butt joint) ...... 171

Figure 93 Residual stress distributions in x, y and z directions measured experimentally in a) {222} and b) {311} hkl planes using neutron diffraction technique for W47 from reactor based diffractometer E3 and calculated using either plane stress assumption, globally applied far-field reference value or local stress free value from reference sample compared to numerically simulated residual stress distributions (AA 2024-T3, butt joint) ...... 174

Figure 94 Residual stress distributions a) across the entire width and in b) x, c) y and d) z directions measured experimentally using neutron diffraction technique for W99 from pulsed spallation source ENGIN-X fitted with Rietveld refinements and calculated using either plane stress assumption, globally applied far field reference value or local stress

xvi free value from reference sample compared to numerically simulated residual stress distributions (AA 2024-T3, butt joint) ...... 176

Figure 95 Residual stress distributions a) across the entire width and in b) x, c) y and d) z directions measured experimentally using neutron diffraction technique for W102 from pulsed spallation source ENGIN-X fitted with Rietveld refinements and calculated using either plane stress assumption, globally applied far field reference value or local stress free value from reference sample compared to numerically simulated residual stress distributions (AA 2024-T3, butt joint) ...... 177

Figure 96 Residual stress distributions a) across the entire width and in b) x, c) y and d) z directions measured using neutron diffraction technique and numerically simulated for single T-joint specimen in {311} reflection, assuming plane stress or globally applied far- field reference value (AA 2024-T3, T-joint) ...... 178

Figure 97 Residual stress distributions a) across the entire width and in b) x, c) y and d) z directions measured using neutron diffraction technique and numerically simulated for triple T-joint specimen in {311} reflection, assuming plane stress or globally applied far field reference value (AA 2024-T3, T-joint) ...... 180

Figure 99 In plane residual stress distributions on the top surface of a butt welded Ti-6Al-4V sheet in the a) x and b) y directions measured using the XRD technique and from FE simulations (Ti-6Al-4V, butt joint) ...... 181

Figure 100 Residual strain distributions in the a) x, b) y and c) z directions for different hkl planes; and the resulting residual stress distributions for d) {112}, e) {201} and f) {103} reflections measured experimentally using the neutron diffraction technique (Ti-6Al-4V, butt joint) ...... 185

Figure 101 strain variation in the x direction in the reference d0 comb sample in the {112} and {201} reflections (Ti-6Al-4V, butt joint) ...... 186

Figure 102 Schematic of the complex thermo-mechanical-microstructural phase transformation coupling interaction (figure taken from [367]) ...... 188

Figure 103 Temperture dependent thermo-physical material properties of AA 2024-T3 used in the finite element model ...... 193

xvii Figure 104 Temperture dependent thermo-mechanical material properties of a) AA 2024-T3 and b) AA 4043 used in the finite element model (figure taken from [386])...... 193

Figure 105 Temperture dependent thermo-physical material properties of Ti-6Al-4V used in the finite element model [387,388] ...... 194

Figure 106 Temperture dependent thermo-mechanical material properties of Ti-6Al-4V used in the finite element model ...... 195

Figure 107 Positions of thermocouples during welding ...... 195

Figure 108 Mesh details of a) butt welded plates, b) T-joint fillet welded plates with one stiffener and c) and three stiffeneres with a user controlled refined mesh ...... 197

Figure 109 3D conical Gaussian type heat source model used for both butt welds and T-joint fillet welds ...... 199

Figure 110 Thermal boundary conditions for butt welded plates and T-joint fillet welded plates showing heat loss due to convection in air, radiation from the surface of the workpiece and conduction between workpiece and the support ...... 202

Figure 111 Mechanical boundary conditions applied during welding and cooling for differernt simulations ...... 206

Figure 112 Temperature contours obtained from thermal analysis when half way through butt welding AA 2024-T3 and comparison of butt weld transverse cross-section geometry through thickness between FE model and experimental macrograph juxtaposed under three different welding conditions ...... 208

Figure 113 Temperature contours obtained from thermal analysis half way through single T- joint fillet welding AA 2024-T3 and comparison of simulated and experimental weld transverse cross-section ...... 209

Figure 114 Temperature contours obtained from thermal analysis half way through triple T- joint fillet welding AA 2024-T3 and comparison of simulated and experimental weld transverse cross-section ...... 210

Figure 115 Numerically calculated thermal histories calibrated using thermocouple measurements at various distances from the weld centre line for butt welded a) W47, b) W99, c) W102 and d) fillet welded specimens (AA 2024-T3, butt joint) ...... 211

Figure 116 Contours of residual stress distribution on the weld transverse cross-section and top surface of butt welded joints in x, y and z directions through thickness under three different mechaincal boundary conditions (AA 2024-T3, butt joint) ...... 213

xviii Figure 117 Residual stress distributions in a) x, b) y, and c) z directions under three different mechanical boundary conditions representing clamping fixtures, perfect constraints around the workpiece edges and welding without any fixtures (AA 2024-T3, butt joint) ...... 214

Figure 118 Contours of out of plane displacements on the top surface of butt welded joints under three different mechaincal boundary conditions (AA 2024-T3, butt joint) ...... 215

Figure 119 a) Cambering and b) angular out of plane displacements under three different mechanical boundary conditions representing clamping fixtures, perfect constraints around the workpiece edges and welding without any fixtures (AA 2024-T3, butt joint) ...... 216

Figure 121 Residual stress distributions in a) x, b) y, and c) z directions under three different sets of welding parameters (AA 2024-T3, butt joint) ...... 218

Figure 122 Contours of out of plane displacements on the top surface of butt welded sheets under three different sets of welding parameters (AA 2024-T3, butt joint) ...... 219

Figure 123 a) Cambering and b) angular out of plane displacements under three different sets of welding parameters (AA 2024-T3, butt joint) ...... 220

Figure 124 Residual stress distributions a) across the entire width and b) in the weld region in all three principal directions, and g) cambering and h) angular distortions as a function of filler metal modelling technique for a single T-joint fillet welded specimens (AA 2024-T3, T-joint) ...... 221

Figure 125 Residual stress distributions a) across the entire width and b) in the weld region in all three principal directions, and g) cambering and h) angular distortions under three differernt mechanical boundary conditions for a single T-joint fillet welded specimens (AA 2024-T3, T-joint) ...... 222

Figure 126 Contours of out of plane displacements on the top surface of single T-joint fillet welded specimens under three different mechanical boundary condition (AA 2024-T3, T- joint) ...... 223

Figure 127 Contours of residual stress distribution and out of plane displacement on the weld transverse cross-section and top surface of a single T-joint fillet welded specimens in x,

xix y and z directions welded using either single or dual laser beams (AA 2024-T3, T-joint) ...... 225

Figure 128 Residual stress distributions a) across the entire width and b) in the weld region in all three principal directions, and g) cambering and h) angular distortions of single and dual beam welding modes for a single T-joint fillet welded specimens (AA 2024-T3, T- joint) ...... 226

Figure 129 Contours of residual stress distribution and out of plane displacement on the weld transverse cross-section and top surface of a triple T-joint fillet welded specimen in x, y and z directions under two different mechanical boundary conditions (AA 2024-T3, T-joint) ...... 227

Figure 130 Residual stress distributions a) across the entire width and b) in the weld region in all three principal directions, and g) cambering and h) angular distortions under two differernt mechanical boundary conditions for a triple T-joint fillet welded specimens (AA 2024-T3, T-joint) ...... 228

Figure 131 Temperature contours obtained from thermal analysis half way through welding Ti- 6Al-4V ...... 229

Figure 132 Comparison of weld transverse cross-section geometry through thickness between FE model and experimental macrograph juxtaposed (Ti-6Al-4V, butt joint) ...... 230

Figure 133 Numerically calculated thermal histories calibrated using thermocouple measurements at various distances from the weld centre line (Ti-6Al-4V, butt joint) .. 230

Figure 135 Residual stress distributions a) across the entire width in all three principal directions, near the weld in b) x, c) y, and d) z directions (Ti-6Al-4V, butt joint) ...... 232

Figure 136 Out of plane displacements after cooling down to room temperature and removing any restraints for three different mechanical boundary conditions (Ti-6Al-4V, butt joint) ...... 233

Figure 137 a) Cambering and b) angular out of plane displacements for three different mechanical boundary conditions (Ti-6Al-4V, butt joint) ...... 233

Figure 138 a) Linear coefficient of thermal expansion and b) unit cell volume for each phase, and transformation induced volumetric strains as a function of temperature [425] ...... 235

xx Figure 139 Modelled equilibrium phase fraction of α and β phases as a function of temperature [424,436,438,439] ...... 237

Figure 140 Time temperature transformation (TTT) diagrams from the literature determined experimentally and calculated using JMatPro [428,429,436,443] ...... 238

Figure 141 Kinetic parameters for diffusion controlled transformations derived from JMatPro TTT diagrams using the interpolating function defined by Kelly for 1 and 95% of transformation [443] ...... 238

Figure 142 Phase fraction of β and α´calculated using the Koistenen Marburger model as a function of a) 푀푠 and b) γ ...... 239

Figure 143 Variation of 0.2% offset yield strength of Ti-6Al-4V with temperature at three different strain rates calculated from JMatPro ...... 241

Figure 144 Time-temperature history of the phase fractions of α, β and αm phases at the weld centre as a function of temperature during a) welding and after b) PWHT at 600°C for 6 hours (Ti-6Al-4V, butt joint) ...... 242

Figure 145 Simulated weld transverse section (20 mm wide) showing the phase fractions of α,

β and αm phases at different stages of welding and after PWHT at 600°C for 6 hours (Ti- 6Al-4V, butt joint) ...... 242

Figure 146 a) RS distributions across the entire width of the workpiece in principle directions, b) x, c) y and d) z stress distributions around the weld, e) angular and f) cambering out of plane distortions under various simulation conditions including phase transformation effect and PWHT (Ti-6Al-4V, butt joint ...... 243

Figure 147 Residual stress distribution in three principal stress directions after welding, on the weld transverse cross-section (20 mm wide) and top surface, with phase transformation (Ti-6Al-4V, butt joint) ...... 244

Figure 148 Y stress distribution in the welding direction viewed from the top surface and close up weld transverse cross-section (20 mm wide) with or without post weld heat treatment (PWHT) as well as either taking into account the effect of phase transformation (Ti-6Al- 4V, butt joint) ...... 244

Figure 149 Y stress distirbution after welding when pre-loaded to 5, 10, 20, 30, 50 and 90% of the room temperature yield stress of the base metal by mechanical tensioning, view from the weld transverse section and top surface (Ti-6Al-4V, butt joint) ...... 247

xxi Figure 150 Residual stress distributions in the y direction as a function of the applied tensioning level, pre-loaded to 5, 10, 20 ,30, 50 and 90% of the room temperature yield sterngth of the base metal (Ti-6Al-4V, butt joint) ...... 248

xxii LIST OF TABLES

Table 1 Main laser sources used for high power laser beam welding [27,78] ...... 20

Table 2 Chemical composition of AA 2024-T3 (Wt. %) ...... 34

Table 3 Chemical composition of Ti-6Al-4V Grade 5 (Wt. %) ...... 34

Table 4 Examination and tests for welds in accordance with acceptance level D specified in BS EN ISO 15614-11:2002 [239] ...... 39

Table 5 Relevant welding quality acceptance criteria from AWS D17.1:2001, BS EN ISO 13919-2-2001 and BS EN 4678:2011 standards (t = thickness) [19,231,240] ...... 41

Table 6 Weld quality assessment criteria from Table 5 applied to 3 mm thick AA2024-T3 as defined by AWS D17.1, BS EN ISO 13919-2 and BS EN 4678 [19,231,240] ...... 43

Table 7 Weld quality assessment criteria from Table 5 applied to 2 mm thick Ti-6Al-4V as defined by AWS D17.1, BS EN ISO 13919-2 and BS EN 4678 [19,231,240] ...... 89

Table 8 Examination and tests for welds in accordance with acceptance level C and D .... 117

Table 9 Differernt sets of welding parameters used for tensile testing macro-tensile welded specimens ...... 124

Table 10 hkl specific E and ν of aluminium following the Kröner Model [352] ...... 161

Table 11 hkl specific E and ν of titanium [353] ...... 161

Table 12 Lattice planes weakly and strongly affected by intergranular strains [354] ...... 162

Table 13 Welding parameters used to perform fibre laser welding experiments on butt welded and T-joint fillet welded sheets ...... 192

Table 14 Mesh details for butt welded and T-joint fillet welded plates ...... 196

Table 15 Calibrated values for heat transfer coefficients and radiation constants ...... 203

Table 16 Calibrated heat source parameters for welding AA 2024-T3 under various welding conditions...... 209

xxiii NOMENCLATURE

α Alpha α´ Martensite αm Martensite and Massive Alpha β Beta γ Material Dependent KM Parameter ε Emissivity εhkl Crystallographic Plane Specific Elastic Strain εxx Direct Strain in xx direction εyy Direct Strain in yy direction εzz Direct Strain in zz direction Total εij Total Strain e εij Elastic Strain p εij Plastic Strain th εij Thermal Strain c εij Creep Strain vp εij Viscoplastic Strain tp εij Transformation Plasticity Strain η Heat Source Efficiency θ Incoherent Stable CuAl2 Phase θhkl Bragg Angle θ0 Stress Free Diffraction Angle θ´ Semi-Coherent CuAl2 Phase θ´´ Semi-Coherence CuAl2 Phase Θ Beam Divergence 휆 Wavelength of Laser νhkl Crystallographic Plane Specific Poisson’s Ratio ρ Density ρabl Ablation Pressure ρg Vapour Flow Pressure ργ Surface Tension Pressure ρhs Hydrostatic Pressure ρhd Hydrodynamic Pressure σ Stefan-Boltzmann Constant σ11 Transverse Stress σ22 Longitudinal Stress σ33 Normal Stress σxx Direct Stress in xx direction σyy Direct Stress in yy direction σzz Direct Stress in zz direction 휔0 Beam Waist AA Aluminium Alloy Ar Argon BM Base Metal BPP Beam Parameter Product BTR Brittleness Temperature Range Cp Specific Heat CET Columnar to Equiaxed Transition CMM Coordinate Measuring Machine CP Commercially Pure CW Continuous Wave

xxiv dhkl Lattice Plane Spacing d0 Stress Free Lattice Spacing DIC Digital Image Correlation DT Damage Tolerance Ehkl Crystallographic Plane Specific Elastic Modulus EDX Energy Dispersive X-ray Spectroscopy eq fα Equilibrium Fraction of α fα Current Fraction of α fα´ Current Fraction of α´ f Current Fraction of αm αm eq fβ Equilibrium Fraction of β fβ Current Fraction of β f Focal Position FE Finite Element FEA Finite Element Analysis FEM Finite Element Modelling FSW Friction Stir Welding FZ Fusion Zone GP Guinier-Preston GPB Guinier-Preston-Bagaryatsky h Planck’s Constant hconv Convective Heat Transfer Coefficient HAZ Heat Affected Zone He Helium HEDB High Energy Density Beam hkl Miller Index HZB Helmholtz Zentrum Berlin ILL Institut Laue Langevin JMAK Johnson Mehl Avrami and Kolmogorov k Thermal Conductivity kβ→α Temperature Dependent JMAK Parameter kβ→αʹ Material Dependent KM Parameter K Strength Index KM Koistinen and Marburger L Total Flight Path Distance LBW Laser Beam Welding LSND Low Stress No Distortion M2 Beam Quality Factor mn Neutron Rest Mass Ms Martensite Start Temperature MSR Mechanical Stress Relieving Treatment n Work Hardening Exponent nβ→α Temperature Independent JMAK Parameter ND Neutron Diffraction Nd:YAG Neodymium Doped Yttrium Aluminium Garnet P Laser Power PM Parent Metal PW Pulsed Wave PWHT Post Weld Heat Treatment qconv Convective Heat Loss q0 Maximum Volumetric Power Density qr Volumetric Power Density qrad Radiative Heat Loss

xxv Q Volumetric Heat Flux r Current Radius of Interior Point r0 Radius of Heat Source at Height z Rw Root Width to Face Width Ratio RS Residual Stress S Incoherent stable CuMgAl2 Phase S´ Semi-Coherent CuMgAl2 Phase S´´ Semi-Coherent GPB2 zones, S11 Transverse Stress S22 Longitudinal Stress S33 Normal Stress SALSA Strain Analyser for Large Scale Engineering Applications SEM Scanning Electron Microscopy SSS Supersaturated Solid Solution tf Isothermal Time to Stage f thkl Time of Flight of hkl Peak Centre * t0 Fictive Time Required to Reach fα(t0,T0) T Temperature T0 Current Temperature Tabs Absolute Zero Temperature Tβ Beta Transus Temperature Tb Boiling Point Tliquidus Liquidus Temperature Tm Melting Temperature Tsolidus Solidus Temperature TEM Transverse Electromagnetic Modes TTT Time Temperature Transformation UTS Ultimate Tensile Strength V Welding Speed Vf Filler Metal Feed Rate WM Weld Metal XRD X-ray Diffraction Yb:YAG Ytterbium Doped Yttrium Aluminium Garnet YS Yield Strength ze Top Surface of the Cone Region zi Bottom Surface of the Cone Region

xxvi 1 INTRODUCTION

The aircraft industry has long recognised the importance of climate protection and the benefits of reducing weight for the production of cost effective and fuel efficient aircraft structures. Reduction of 20% weight would save about 10% of fuel and even further reduction of 30% weight would save around 10% of operation cost [1]. One of the key ways to achieve this is by using light weight alloys such as high strength titanium alloy Ti-6Al-4V and aluminium alloy AA 2024-T3 for their superior strength to weight ratio. The 2xxx series aluminium alloys are widely used for lower wing skin and fuselage of aircraft structures, and machine parts such as bolts and rivets. The 6xxx series alloys are mainly used in extruded parts and the 7xxx series alloys where high mechanical strength is required [2].

Titanium alloys are often used primarily as a steel replacement for weight savings or in critical applications where there is a space limitation and the strength of aluminium alloys is not an option. The 2xxx series alloys however, are known to be difficult to weld due to their susceptibility to cracking and high surface reflectivity during welding. Titanium alloys on the other hand is readily weldable and less sensitive to cracking but because of the high affinity for oxygen, nitrogen and hydrogen, embrittlement can arise from contamination if not adequately protected. Welding defects such as cracks and porosities pose a significant risk to the welded parts since they tend to weaken the structure by introducing stress concentrations around the defects. These defects often occur when the power density of the welding process is not high enough or due to residual stresses [3]. Furthermore, steep temperature gradients in the fusion zone trigger vaporisation and loss of some alloying elements in aluminium alloy, which may degrade its mechanical properties. The formation of porosity during welding increases the tendency for the welded structures to fail or even fracture [4,5]. However, these barriers can be mitigated by optimising the power density and heat input, both of which can be controlled by the laser source parameters (e.g. laser power, welding speed, focal position and etc.). In this way, more stable and efficient laser welding mode can be achieved.

Traditionally, riveting is chosen over welding to fabricate aircraft structures in the aircraft industry. The technology is now considered well established with many experiences and numerous research to understand the performance of the conventional riveted structures such as tensile, formability, damage tolerance (DT) requirements of fatigue, fracture and residual stresses, and other static and dynamic properties as well as corrosion behaviours. Future aircrafts, however, have challenging design objectives to reduce cost and weight of the structures while at the same time, become faster and safer as well as to reduce carbon emissions.

1 Recent developments in welding technologies have led to innovative use of material forms and advanced joining methods for example, high power laser beam welding (LBW) and friction stir welding (FSW) for welding and machining lightweight alloys to obtain highly integrated structures [6], which is another way of reducing weight and fabrication costs. For example, replacement of rivets by CO2 laser beam welding and friction stir welding (FSW) have been applied successfully to manufacture various commercial aircraft components such as skin- stringer curved panels in fuselage and wing structures with weight and cost savings, each of around 15% [7]. A few examples include the Airbus A340 and A380 [5] and the Comac C919. However, the effects of using advanced welding processes to produce welded structures on their mechanical properties, especially the damage tolerance behaviour, have only been investigated by a limited number of studies in recent years [8,9].

Integral welded structures unlike differential riveted structures, are subjected to significant changes in the microstructure, metallurgical state and associated mechanical properties of the materials in the welded joints [10,11]. Such changes can result in residual stress, distortions and defects formation in the structure, thus significantly influencing the structural integrity and service life of the lightweight alloy products. Specific features of the lightweight alloys in both welding and joining processes need to be clearly understood and defined by the designers for the selection of materials at the early stage of design and by manufacturers for process control at the manufacturing stage. In order to maintain structural integrity of welded structures for the whole service life of the structure, the relationship between welding process, weld properties and performance of the structure needs to be fully assessed [10]. As a result, it is necessary to establish comprehensive relationships of materials, welding processes, and microstructure and mechanical properties of laser welded joints [12].

Due to a deep penetration keyhole formed within the weld material and its large depth to width ratio [13], laser welding is considered superior to conventional arc welding processes especially for aircraft applications. The high power density and associated keyhole characteristics can also lead to welds with narrow heat-affected zones, and reduced residual stresses and distortion. In certain applications, high power density beam welding can offer both high quality and cost-effectiveness when compared to other welding methods [14]. There are a number of different laser welding systems available in the laser welding industries, from the well adopted CO2 and Nd:YAG lasers to the newly developed fibre, disc and diode lasers. Rapid improvements in the laser technology have resulted in considerable cost and weight savings, meanwhile enhancing structural rigidity and welding efficiency. Although fibre laser has been available since the 1960s, higher power fibre lasers suitable for welding have only been available for the past five years [15,16]. The power range is now available from several hundred watts to over 10 kW, exceeding that for both CO2 and Nd:YAG lasers. The operating

2 wavelength of fibre laser is about 1.07 µm [17] which is similar to the Nd:YAG laser. More importantly, the overall efficiency of fibre laser is much higher than the other two (almost double that of the CO2 laser), reaching around 25% [17]. This gives rise to a narrower weld zone, lower deformation or distortion, which represents improved quality, high precision and good flexibility. As the two dominating CO2 and Nd:YAG lasers have their limitations, there is a good chance for fibre laser to establish a firm market position in the laser welding industry.

The work presented in this thesis is a part of the research program funded by Aviation Industry Corporation of China - Beijing Aeronautical Manufacturing Technology Research Institute AVIC-BAMTRI with the collaboration of Imperial AVIC Centre for Structural Design and Manufacture. Welding was performed at BAMTRI on thin sheets of AA 2024-T3 and Ti-6Al-4V using its own YLS-5000 and YLS-10000 kW Class Ytterbium Fibre Laser Systems. Fundamental academic research on characterisation of fibre laser welding components made from these two alloys was conducted at Imperial College London. The work comprised analysing material behaviour, identifying the effects of welding parameters on weld morphology and properties of welded joints, experimental and numerical modelling of welding processes, and performing mechanical testing and residual stress measurements. The work will guide BAMTRI in its own future applications research.

1.1 Aims and Objectives The overall aim of this research was to investigate welding of aluminium alloy 2024-T3 and titanium alloy Ti-6Al-4V using a high power fibre laser for applications in the aircraft industry. It involved establishing an understanding of the influence of welding parameters on microstructural change, welding defects, mechanical properties, and the characteristics of heat affected zone (HAZ) and weld metal (WM) of fibre laser welded joints. Also, the effects of welding parameters on thermal stresses and the resulting residual stresses and distortions in the welds were investigated. The following objectives were set to achieve the aim.

 To optimise welding parameters to produce full penetration, bead on plate, butt and fillet welds in thin sheets of 3 mm AA 2024-T3 and 2 mm Ti-6Al-4V of a quality acceptable for aircraft structures according to the specifications provided in the European standards ISO 13919-1 [18], ISO 13919- 2 [19], and the American standard AWS D17.1 [20].  To review existing laser welding processes including fibre laser welding, the principles of laser keyhole welding, origins and mechanisms involved with the formation of welding defects, and published work on the subject of welding aluminium and titanium alloys and their applications in the aircraft industry.

3  To perform experimental testing to determine the relationship between the welding process and the heterogeneous microstructural and thermo-mechanical properties around welded joints.  To develop thermo-mechanical FE models on Abaqus to simulate fibre laser beam welding of AA 2024-T3 and Ti-6Al-4V with experimental calibrations to describe residual stresses and distortions in butt welded and fillet welded test plates.  To research into available inspection methods capable of monitoring residual stress across weld cross-sections and compare predictions obtained through numerical techniques with experimental data and to interpret the findings.

1.2 Structure of the Thesis Chapter 1 outlines the context and the background of the research work. The project aims and objectives are described. Chapter 2 reviews the laser beam welding processes currently being used in the aircraft industry, their limitations and the issues and problems in laser beam welding of aluminium and titanium alloys for aircraft applications. Published work on the development of fibre laser welding technology, its advantages and disadvantages, and its potential application in the aircraft industry are discussed. Chapter 3 provides a detailed description of the materials used in this study, in addition to the experimental methods used to characterise the laser welds. The influence of welding parameters including design parameters, technological parameters and metallurgical parameters on weld formation, microstructure in different regions of the weld and defects is described. Chapter 4 describes the procedures for mechanical testing of welded joints and analyses the results of experimental tests on micro-hardness and tensile properties of welded joints. The experimental results are used for the validation of numerical models in Chapter 5. Chapter 5 presents the development of laboratory scale numerical models to simulate fibre laser beam welding of AA 2024-T3 and Ti-6Al-4V under industrial loading and boundary conditions. Experimental study on the measurements of welding thermal cycle and welding induced residual stresses and distortions were conducted to ensure the reliability of the numerical models. Chapter 6 summarises the conclusions and provides suggestions and recommendations for future work. Appendix describes the subroutines used in numerical models to define the fibre laser welding heat source and solid state phase transformations associated with Ti-6Al-4V during welding and post weld heat treatment (PWHT).

4 2 LITERATURE REVIEW

2.1 Introduction Applications of laser beam welding process for light alloys such as AA 2024-T3 and Ti-6Al-4V are increasing in the aircraft industry, where high performance lightweight structures are essential. A significant amount of research has been conducted to understand the microstructural and mechanical properties of aluminium and titanium welds. However, the results achieved so far, based on the most relevant prior publications, were largely material, welding process and application specific, and therefore cannot be generalised and is expected to vary for different conditions. This chapter reviewed the structure and properties of AA 2024- T3 and Ti-6Al-4V and their applications in the aircraft industry; basic concepts in the characteristics of lasers and mechanism of laser beam welding; different types of lasers with specific emphasis on high power lasers suitable for welding process including CO2, Nd:YAG and fibre lasers; and finally the progress in laser beam welding of aluminium and titanium alloys relevant to this investigation.

2.2 Structure and Properties of AA 2024-T3 Aluminium alloys are known for their unique high strength to density ratio and stiffness superior to steel, making them suitable and popular for airframe construction. AA 2024-T3 belongs to the Al-Cu-Mg group of the 2000 series heat treatable aluminium wrought alloys in which copper and magnesium are the primary alloying elements. The properties of the 2024 alloy are enhanced through the addition of higher levels of alloying elements and controlled heat treatments which consist of solution heat treatment, quenching, cold working and then thermal age hardening. The T3 temper involves solution heat treatment, cold working and then natural aging to a substantially stable condition. In the 2024 alloy, copper substantially increases strength and promotes precipitation hardening but also reduce ductility and corrosion resistance. Magnesium is added to increase strength and hardening by natural aging at room temperature. Precipitation hardening of the 2024 alloy involves a four stage transformation sequence for the aging of Al-Cu alloys given in Equation 1 [21], as well as that for the aging of Al-Cu-Mg alloys due to the secondary addition of magnesium in its composition, given in Equation 2 [2,22].

SSS → GP zone → θ´´ (CuAl2) → θ´ (CuAl2) → θ (CuAl2) Equation (1)

SSS → GPB zone → S´´ (GPB2) → S´ (CuMgAl2) → S (CuMgAl2) Equation (2) where SSS stands for supersaturated solid solution, GP for Guinier-Preston and GPB zone for Guinier-Preston-Bagaryatsky zone [2,21,22].

5 The solid solution strengthening and precipitation hardening is mostly achieved by the heterogeneous nucleation of predominantly semi-coherent S´ (CuMgAl2) phase and a fewer semi-coherent θ´ (CuAl2) phase on dislocations by diffusion. The contribution to strength from the GPB zones is relatively small [2]. A trace amount of silicon in the 2024 alloy enhances the nucleation of the semi-coherent phases and increase the stability of GPB2 zones but also induce the formation of coarse Mg2Si precipitates which reduce the fracture toughness.

The 2024 alloy is often used in the T3 temper as naturally aging alone provides high ratios of tensile to yield strength particularly at elevated temperatures, ductility, fracture toughness and fatigue resistance but increase susceptibility to stress corrosion cracking. Still, cold working after solid solution and before aging improves strength and stress corrosion resistance of the 2024 alloy by increasing the number of nucleation sites for age hardening precipitates [23]. Although it has a good response to artificial aging if cold worked prior to aging, artificial aging is a less preferred approach as its fatigue strength and fracture toughness may decrease even if yield strength increases. In other words, fatigue and fracture performance in the naturally aged condition is superior to that in the artificially aged condition. In the case where it is difficult to thermally relieve high residual stress induced from the quenching process to obtain high strength, for example, it can be stress relieved by plastically stretching in the rolling direction from 1.5 to 3% strain and is designated as the T351 temper.

Aluminium alloys possess a number critical inherent characteristics such as high thermal conductivity and thermal expansion coefficient, high reflectivity to laser, low boiling point elements, high hydrogen solubility in liquid state and aluminium oxide films which make welding of aluminium alloys a rather complex problem. The high reflectivity of aluminium to light of around 95% at room temperature due to high electrical conductivity, acts like a mirror and reflects a large proportion of the beam energy and therefore, reduces its weldability. The loss of energy and rapid heat dissipation due to the high thermal conductivity mean that large heat inputs or high power density are required to weld aluminium and increase sensitivity to fluctuations in heat input by the welding process. Reflectivity is also a function of the wavelength of the light source through the dispersion relation of its refraction index as shown in Figure 1 for the relationship between reflectivity of different materials and wavelength [24]. It can be seen that there is a considerable difference in the reflectivity of aluminium between wavelengths of around 1 μm, similar to that for Nd:YAG and fibre lasers, and 10 μm for CO2 lasers, where the reflectivity is lower with a value of around 0.85 at 1 μm and greater at 10 μm, close to 0.95, indicating that almost all the incident radiation is reflected. Reflectivity and absorptivity are both dependent on the wavelength and temperature. The absorptivity of aluminium was reported to increase from less than 0.1 at room temperature to 0.4-0.5 at its melting point and as high as 0.9 at its vaporisation point [25].

Figure 1 Reflectivity of aluminium and other metals as a function of laser wavelength for normally incident light on a flat surface at room temperature [26,27]

While pure aluminium is not susceptible to hot cracking due to the absence of low melting point eutectic at the grain boundary, the addition of alloying elements such as Cu, Mg and Si to the 2024 alloy makes its composition crack susceptible. As a result, it is considered to have poor weldability because of its high solute content, resulting in a wide solidification temperature range and increased tendency to form low melting constituents along grain boundary, known as the eutectic composition with lower freezing points than the base metal; and also due to high coefficient of thermal expansion and large solidification shrinkage. Hot cracking is a high temperature cracking mechanism and it can refer to either solidification cracking which occurs intergranularly along the grain boundary of the weld metal during solidification due to the stresses acting on the segregated low melting point constituents [28], or liquation cracking which occurs in the coarse heat affected zone along the fusion boundary as a result of grain boundary liquation of the low melting point constituents. Liquation cracking is less frequently observed in laser welds due to the low heat input and small size of the HAZ and partially melted zones [29]. Unlike steel, aluminium is not susceptible to cold cracking caused by hydrogen embrittlement [30]. Different theories to explain the mechanism of solidification cracking in welding was reviewed by Cao et al. [31], including the strain theory, the brittleness temperature range theory (BTR), Borland’s generalised theory and critical speed theory. The 2024 alloy has a range of freezing temperatures as a result of the presence of alloying elements so the solidification strains are proportional to the temperature interval over which solidification occurs or the coherence (mushy) range between the formation of coherent interlocking of dendrites and the solidus temperature; and mechanical properties of the alloy in the temperature range of formation of cracks. The effective solidus temperatures are depressed by rapid non-equilibrium solidification during welding. During solidification, the areas with higher freezing points solidify first from the fusion boundary towards the weld centreline. The low melting point constituents are rejected by the formation of a coherent

7 dendrite interlocking solid network and the remaining liquid along grain boundaries form continuous thin inter-dendritic liquation films which persist down to low temperatures. In the liquid film strain concentration theory, a crack is formed when the liquid films cannot withstand the thermal shrinkage strains developed in the solidifying weld metal and the supply of residual eutectic liquid available to fill the incipient cracks caused by the shrinkage strains between grains during freezing is insufficient [32].

Hot cracking sensitivity of the 2024 alloy can be reduced by controlling the solidification process during welding by optimisation of welding parameters and avoiding the crack sensitive composition in the weld metal through the use of a suitable filler metal with different chemistry and a lower solidification point than that of the base metal [33]. Fine equiaxed dendritic structure with sufficient liquid between grains can increase solidification crack resistance, by decreasing the coherent temperature range thus allowing it deform more easily than columnar dendritic structure. High power density of laser beams significantly reduce thermally induced strains and produce fine grains due to low heat input and rapid cooling rate. Fine grains lower the crack susceptibility by distributing the low melting point constituents over a larger grain boundary area and relieve local shrinkage strains during freezing [34]. While increasing the welding speed leads to a faster cooling rate and less time spent in the hot cracking temperature range, high thermal shrinkage strains rapidly develop as well and results in a higher crack initiation rate and shorter time to heal initiated cracks. Therefore, higher heat input and lower welding speed are used to decrease cooling rate and shrinkage stresses in order to reduce solidification cracking [31,35,36]. On the other hand, addition of grain refiners such as titanium and zirconium can refine the solidification structure and reduce hot cracking. The hot cracking sensitivity of the 2024 alloy in terms of copper content is low as it is greater than the amount required for the peak sensitivity but the small amount of Mg which depresses the solidus and extends the coherence range, is close to the maximum sensitivity range. By using a filler alloy such as 4043 and 4047 with higher percentages of silicon, the composition can be modified to avoid solidification cracking and also the low freezing range of 4043 ensures that the weld is the last to solidify and allows the base metal to solidify prior to the weld.

Porosity is also a particular concern when laser beam welding aluminium alloys [12] but its effect is less critical than solidification cracking as it can be tolerated up to levels defined in various laser welding related standards. The presence of porosity reduces the weld cross- sectional area and when present in sufficient size and quantity, it can cause stress concentration and lead to a reduction in transverse tensile strength, fatigue resistance and ductility of the weld metal [37,38]. There are two main types of porosity formed during keyhole laser welding aluminium alloys, which are classified according to the size of pores, where

8 macro-porosity is caused by keyhole instability and micro-porosity is mostly related to hydrogen in the weld pool. Large and irregular shaped keyhole porosity is mainly observed in partial penetration welds due to keyhole collapse and instability, where molten metal is unable to feed steadily the keyhole cavity behind the moving laser beam and so metal vapours including low boiling point alloying elements, oxides, shielding gas and contaminants become trapped at the root of the weld [39]. Aluminium alloys are also very susceptible to hydrogen related porosity because hydrogen solubility in aluminium is an exponential function of temperature, exhibiting almost 20 times higher solubility of 0.65 ml/100g in liquid aluminium during welding compared with 0.034 ml/100g in solid aluminium [35], and even 2 ppm of hydrogen in the molten pool is enough to produce porosity in aluminium welds [40]. A large amount of hydrogen is absorbed in the weld pool at high temperatures but as it solidifies, the solubility decreases and as a result, retains much less hydrogen and dissolved hydrogen in excess of the effective solubility limit is rejected at the solid-liquid interface to form gas bubbles.

Small hydrogen pores of spherical shape are produced in the weld metal when there is not enough time for these gas bubbles to escape out from the solidifying weld pool and instead become trapped in the weld metal due to the rapid cooling rate [35]. The amount of porosity depends on the solidification rate so increasing the laser power or decreasing the welding speed will reduce the cooling rate and allow gases to escape from the weld pool. It is also possible to reduce porosity by further increasing the cooling rate as it means that the weld pool solidifies before bubbles can grow. The sources of hydrogen include surface contaminants such as oil, moisture, and dirt on the workpiece and filler metal; inclusions such as intermetallic compounds, aluminium oxides and hydroxides; atmospheric gases like nitrogen and oxygen; and type of shielding gas and flow rate. It is therefore, essential to remove these sources by rigorously cleaning the joints using different methods such as chemical cleaning and steel wire brushing immediately prior to welding, and supplying a stable and effective shielding gas shroud over the molten weld pool during welding.

Welding the 2024-T3 alloy is associated with loss of strength and ductility in the FZ and the HAZ. The loss of strength or softening in the FZ is caused by high temperatures of the welding thermal cycle which induce dissolution of strengthening precipitates in the matrix. Due to the sufficiently high temperatures and fast cooling rates during solidification, solution treatment takes place and some aging occurs after welding so the strength of the unaffected BM can be partially regained by post-weld hardening [41]. In addition, loss of low boiling point alloying elements such as magnesium by vaporisation due to high equilibrium vapour pressure results in loss of precipitation strengthening and causes reduction in yield strength, ductility and hardness, but increase crack susceptibility [42–45]. The vaporisation rate can be lowered to reduce elements losses by welding at higher welding speeds and moderate laser powers [46].

9 The loss of strength in the HAZ depends on the peak temperature experienced during welding, where overaging by coarsening of the strengthening precipitates and transformation to non- strengthening precipitates occurs at lower peak temperatures, while dissolution of strengthening precipitates and partial melting of the grain boundaries take place at higher peak temperatures which remain in the solution annealed condition unless the cooling rate is sufficiently fast to cause reprecipitation by natural aging.

2.2.1 The Use of AA 2024-T3 in Aircraft Applications Aluminium is a popular material in the aircraft industry due to its light weight, moderate cost and good mechanical properties including specific stiffness, strength and ductility, and good fracture toughness. It accounts for 60-80% of the structural weight of commercial aircrafts and 40-60% for military aircrafts [47]. The two main series of aluminium alloys used for aircraft structures are the 2000 and 7000 series alloys [48,49]. The 2000 series AlCu-Mg alloys such as 2024 are primarily used in tension dominated structures, whereas, the 7000 series Al-Zn- Mg alloys such as 7075 are used for compression dominated structures. Although the 7000 series alloys have higher strength than the 2000 series alloys, they have lower damage tolerance (DT) and stress corrosion resistance so their use is restricted to compression limited structures like upper wing structures where fatigue crack growth is less critical [50]. AA 2024- T3 has been used in aircraft since the 1930s mainly for fatigue critical structures where predominantly tensile loading conditions [51]. It is mostly used in the cold worked and naturally aged T3 temper for fuselage skin, frames and bulkheads as it leads to high stiffness, strength to density ratio and DT [51]. DT, in addition to static strength is an important property relating to the ability to operate safely over its lifetime to the designed static and dynamic in-service loading conditions and sustain defects and other forms of damage until repair. The 2024-T3 alloy with high DT, has a combination of high fracture toughness, low cycle fatigue strength and resistance to fatigue crack growth and corrosion [52] and therefore, making it suitable for fuselage structures where good static strength, fatigue and fracture resistance are needed due to repeated cycles of pressurisation and depressurisation of the cabin every flight. More recently, other 2000 series alloys such as 2224, 2324 and 2524-T3 were developed to produce improved tensile strength, fracture toughness and fatigue crack growth resistance compared to the 2024-T3 due to lower iron and silicon contents which reduce the volume fraction of brittle constituent phases [53,54]. While it is useful to develop new alloys with better performance, it is also important to extend the availability of existing alloys.

Although initially attempts were made to weld AA 2024-T3 aircraft structures, riveting became the dominant joining method due to the difficulties associated with fusion welding of the 2024 alloy [55]. Majority of the aircraft fuselages at present are joined by riveting, which is estimated to consume around 40% of the total manufacturing time of the aircraft structure [56]. On

10 average, one million rivets are used in the fuselage structure of a typical commercial aircraft at a processing rate of 100-400 mm/min or 4-5 rivets per minute, and is associated with added weight from the rivets and overlapping or strengthening joints. In addition, rivet holes can act as source of fatigue cracking and corrosion [57]. Alternative manufacturing methods to riveting have been investigated by aircraft manufacturers for many years, driven by environmental and commercial factors to improve fuel efficiency, reduce emissions and costs through weight savings [58], and increase productivity since manufacturing represents 95% of the cost of a commercial airframe structure which in turn is around 20% of the total aircraft cost [48]. Laser welding is considered as one of the alternative methods to rivets and fasteners for applications in fuselage and wing structures due to its high quality, low heat input, high power density, fast welding speed, narrow heat affected zone and low distortion. The use of lighter integral welded structures increases productivity at processing rates up to 10 m/min which is 20 times faster than riveting [57] and reduces the manufacturing cost by about 20% and saves up to around 10% of the total weight of the aircraft relative to riveting [55]. However, welding is a permanent joining method and the heat from welding can affect the microstructure and mechanical properties of the welded joint and induce distortion and residual stress due to local heating during welding. Unlike riveted structure, it is difficult to arrest cracks in integral welded structures once they start to propagate, especially those in the weld metal such as solidification cracks and porosities. The lower strength of the weld metal leads to plastic strain localisation within the softer weld metal and consequently, increase the crack tip constraint and affect the fracture performance of the welded structure [59].

The adoption of lasers for welding aluminium aircraft structures has been slow and limited, initially due to limited laser power and beam quality which made deep penetration keyhole welding difficult [12] and also because of high equipment costs [57]. High power CO2 lasers with sufficient power density to produce a stable keyhole welding has been available for many years and a range of other laser sources also became available more recently which are as good as or even better than the CO2 laser for welding applications. Nd:YAG and fibre lasers are the two other laser sources which can be used for the welding of aluminium aircraft structures. They both have a wavelength of around 1 μm which is around ten times shorter than that of the CO2 laser, allowing fibre optic delivery of the laser beam and improved laser- material interaction due to more efficient absorption of the laser light by aluminium. The capability of the Nd:YAG laser for welding 2024-T3 has been demonstrated in a limited number of publications but very little has been published on the fibre laser which only became commercially available at high powers comparable to those of the CO2 lasers and exceeding those of the Nd:YAG lasers in the late 2000s [48]. Fibre laser has better beam quality than the other two laser sources and power efficiency at least three times greater than the CO2 lasers,

11 which allows faster welding speeds to be utilised. However, the use of fibre laser for welding AA 2024-T3 has not been well established yet and therefore, still requires extensive amount of work to be done in order to enhance its capabilities and produce high quality welded joints free of crack and with minimum level of porosity [60,61].

Even though the 2024-T3 alloy is the preferred material for lower fuselage skin, it is replaced by more weldable 6000 series aluminium alloys because it is considered to be difficult to weld due to its crack sensitivity and susceptibility to porosity. However, as the 6000 series alloys exhibit lower strength compared with the 2024-T3 alloy, thicker sections have to be used at the cost of increased weight to maintain the strength and stiffness of structures equivalent to those manufactured using the 2024-T3 alloy [59,62]. The lower fuselage panels are subjected to a combined loading condition of compression, shear and hoop stress [63]. More than 200 lower fuselage skin panels and bulkhead panel of the Airbus A318, A380 and A340-600 HGW with laser beam welded skin-stringer joints, which meet the requirements of the aircraft industry, were produced from 6013 and 6056 aluminium alloys using 3.5 kW dual CO2 laser beams and 4047 filler metal to prevent hot cracking and porosity [58,64]. However, it has been found that the same material combination of 6000 series alloys as that used in the lower fuselage panels do not provide sufficient strength advantages compared to riveted structures for upper fuselage applications in terms of its damage tolerance, and the different loading condition of tension and longitudinal hoop stress. Thus, higher residual strength and improved fatigue behaviour are required [63]. The use of welded 2024-T3 alloy, on the other hand, would be suitable for the upper fuselage panels because it has sufficient fatigue resistance and damage tolerance which inhibit fatigue crack growth and raise the limit load bearing capacity [65].

2.3 Structure and Properties of Ti-6Al-4V There are four types of titanium alloys on the basis of their microstructure, including commercially pure (CP) titanium, alpha alloys, alpha-beta alloys and beta alloys. CP titanium and alpha titanium alloys are non-heat treatable and can only be solid solution strengthened or cold worked by alpha stabilising elements such as aluminium, silicon and oxygen, so they have relatively low strength, good ductility, creep strength and corrosion resistance. Beta or near beta alloys are heat treatable, weldable and have high tensile and fatigue strength but have low stiffness and ductility. Alpha-beta alloys consist of a mixture of alpha and beta phases and Ti-6Al-4V is the most frequently used alpha-beta alloy, containing 6 wt% aluminium as the alpha stabiliser and 4 wt% vanadium as the beta stabiliser. They are heat treatable, weldable, considerably stronger than alpha alloys and have better creep strength and ductility than beta alloys [66,67]. Titanium undergoes allotropic transformation where the crystal structure changes depending on the temperature and chemical composition. The

12 hexagonal close packed (HCP) alpha phase forms below the beta transus, the lowest temperature at which a fully beta phase can exist, whereas, the body centre cubic (BCC) beta phase only forms above the beta transus. However, the addition of controlled amount of beta stabilising alloying elements such as chromium, molybdenum, tin, niobium and vanadium increases strength, ductility, toughness, hot workability and promotes the formation of the beta phase below the beta transus. On the other hand, the addition of alpha stabilisers such as aluminium, oxygen, nitrogen and carbon increases strength and extends the alpha phase field to higher temperatures [67,68].

Ti-6Al-4V can be strengthened using different heat treatments depending on the required mechanical properties, some of which includes mill annealing (MA), recrystallization annealing (RA), beta annealing (BA), and solution treatment and aging (STA). Mill annealing is the most common heat treatment which provides an optimum combination of high tensile strength, good fatigue strength, and moderate fracture toughness and resistance to fatigue crack growth. Recrystallization annealing and beta annealing are used to increase fracture toughness and resistance to fatigue crack growth but tensile and fatigue strength are reduced. Recrystallization annealing is used for fracture critical applications and beta annealing is used for damage tolerance applications. Solution treatment and aging provides the maximum strength around 50% greater compared with the annealed condition but significantly reduces ductility so it is used for high strength applications [67–69]. Ti-6Al-4V base metal used in this investigation was in the mill annealed condition as it has good weldability in this condition, with equiaxed alpha grains with colonies of Widmanstätten alpha plates at grain boundaries and untransformed residual beta phase [70].

Its microstructure and properties resulting after welding depends on the cooling rate and peak temperature reached during welding. Cooling from above beta transus at slow cooling rates (furnace) results in the same microstructure as the base metal via a diffusion controlled nucleation and growth of coarse alpha phase on the prior beta grain boundaries and a colony of Widmanstätten plates within grains [70]. For higher cooling rates (air) but not fast enough for diffusionless transformation, fine plate like acicular alpha phase is formed. At even faster cooling rates (quench) from temperatures above the martensite start temperature, diffusionless martensitic transformation from beta phase to a metastable fine needle like or acicular hcp martensite, α´, occurs [67]. The martensite start and end temperatures depend on the alloy composition or the percentage of beta stabilisers, which in Ti-6Al-4V is low enough to keep the martensite start temperature above the room temperature. As a result, quenching from the beta transus does not retain the metastable beta phase but instead produces martensite, which leads to an increase in strength and some hardening effect but lower ductility due to the phase transformation [66]. Post weld heat treatment (PWHT) aging is often

13 performed on Ti-6Al-4V welds at elevated temperatures to stress relieve the residual stresses developed during welding and to increase strength [66]. During the heat treatment, the martensite decomposes back to alpha and beta phases of equilibrium composition at the aging temperature [71].

Ti-6Al-4V has good weldability due to its low beta stabiliser content. Increasing the beta content makes it more difficult to weld as it leads to microstructures of lower toughness and ductility and become more susceptible to liquation cracking. It is frequently welded using electron beam welding under high vacuum condition for aircraft applications. Fibre laser welding can offer similar welding performance to electron beam welding without involving the use of vacuum through the atmosphere and provide a very flexible fibre optic delivery of laser beams to the workpiece even in hard to access areas at high power density and fast welding speeds. However, the main problem when welding Ti-6Al-4V is its high affinity for hydrogen, oxygen and nitrogen from shielding gas, atmosphere and the workpiece at elevated temperatures above around 500 °C, where contamination of the weld metal in fusion zone and heat affected zone can result in embrittlement and also formation of porosity [67]. It is therefore, necessary to ensure that the weld area is cleaned prior to welding by degreasing and stainless steel wire brushing, and protected from the surrounding atmosphere by the use of top, root and trailing protective inert gas shielding of the weld pool and the adjacent HAZ during welding [72]. Ti-6Al-4V has low solidification and liquation cracking susceptibility as its solidification temperature range is small and so residual low melting liquid film is not formed. According to Inoue and Ogawa [73], micro-segregation of beta stabilising elements, vanadium and iron at the dendrite boundaries during solidification in Ti-6Al-4V is insignificant due to high solute elements back diffusion rate and subsequent rapid dispersion of micro-segregation in solid [73]. For this reason, it is common to weld Ti-6Al-4V autogenously or with filler metal of the material.

2.3.1 The Use of Ti-6Al-4V in Aircraft Applications Commercially pure (CP) titanium and Ti-6Al-4V are the two main types of titanium and titanium alloys used in aircraft. The CP titanium is unalloyed and it is often used in sections where a good corrosion resistance is required but structural strength is not a critical factor. Titanium alloys have excellent corrosion resistance due to the formation of thin titanium oxide films which are resistant to most corrosive agents so they perform better than high strength aluminium alloys in terms of stress corrosion cracking and exfoliation [66,67] .The Ti-6Al-4V titanium alloy is the most commonly used titanium alloy and is much stronger than the CP Ti so it is ideal for high performance aircraft applications, where structural strength is important and the relatively high material cost can be justified. It has moderate density, high specific strength, excellent corrosion and fatigue resistance, fracture toughness, creep strength,

14 weldability and is heat treatable [74]. Its density is slightly greater than aluminium alloys but around half of that of stainless steel and nickel alloys [66]. Different strength levels comparable to the strength of steels can be obtained by heat treatment so it has one of the highest strength to weight ratio even greater than aluminium alloys [75]. Unlike aluminium alloys, Ti-6Al-4V offers improved resistance to heat as it retains good mechanical properties at elevated temperatures up to around 500-600°C for which other aircraft materials such as aluminium alloys, magnesium alloys and composites cannot be considered [67]. On the other hand, Ti- 6Al-4V has relatively low specific stiffness, similar to that of aluminium alloys but significantly lower than steels, so there is no benefit in using this alloy for high stiffness [67]. Also, the main disadvantage of titanium alloys is the high raw material cost. For that reason, the use of titanium alloys in the aircraft industry for structural applications developed slowly because of their cost relative to aluminium alloys, yet, are still finding increasingly greater preference over aluminium alloys and stainless steel [71] due to demands for lighter, faster and more fuel efficient aircrafts, reduced emissions and the need to replace existing fleets for financial and environmental reasons [66]. The key drivers for using Ti-6Al-4V in aircraft applications include weight savings as a replacement for steels and nickel based superalloys, higher operating temperature performance and stability than aluminium alloys, higher specific strength in a limited space where the strength of aluminium alloys or steels is insufficient, corrosion resistance and composite compatibility [74]. Such elevated temperatures occur in aircrafts that fly at higher speeds due to aerodynamic frictional heating of the fuselage skin as well as in the compressor sections of turbine engines where high specific strength is needed.

Ti-6Al-4V aircraft structures are often produced by machining from solid billet or forged. For parts which have a complex shape, poor formability at room temperature means that it is unavoidable to process at high temperatures. The production costs are high and the parts have high buy to fly ratios, which is the weight ratio between the raw material and the finished component so the economic benefit of using Ti-6Al-4V is reduced. It is however, possible to significantly reduce the buy to fly ratio by manufacturing near net shape components by welding, which also removes the need for mechanical fasteners or adhesive bonding. Ti-6Al- 4V is readily weldable and its low thermal conductivity and high absorption of infra-red light as shown in Figure 1 increases its weldability. While electron beam welding is mostly used to weld Ti-6Al-4V due to its high affinity for hydrogen, oxygen and nitrogen at elevated temperatures, using fibre lasers has the advantages of increased welding process flexibility by fibre optic delivery of laser beams and making operation under vacuum condition unnecessary [66]. As long as it is completely shielded against contamination from the atmosphere during welding to avoid embrittlement of the weld metal, fibre laser welding can be used to produce Ti-6Al-4V aircraft components at increased productivity and reduced costs

15 [27]. Therefore the relationship between fibre laser welding process, and microstructure and properties of Ti-6Al-4V must be investigated [54].

Airframe Engine

Fe CFRP 8% 19% Ti Ni 37% Ti 40% 7%

Al 66% Fe Al 15% 3% CFRP 5%

Figure 2 Structural weight ratio of different materials in commercial aircraft airframe and engine [74]

Figure 2 shows that the main aircraft application of titanium alloys is gas turbine engine, where it accounts for over one third of the structural weight of the engine, besides nickel based superalloys. The initial use of Ti-6Al-4V before 1980 was relatively low at 3-5% of the structural weight, mainly limited to blades and discs in turbines and compressors due to its excellent creep resistance at high operating temperatures up to 350°C and characteristics of superplastic forming and diffusion bonding [67]. Its usage in aircraft structures steadily increased over many years in a wide range of structures including aircraft fuselages to prevent fatigue crack growth, airframes, landing gear components, on-board kitchen and toilet flooring, exhaust shrouds, firewalls, engine nacelles, engine supports, hydraulic tubing, wings, wing boxes, undercarriage parts, cockpit window frames [67,71]. For example, the Boeing 707 commercial aircraft only contained 80 kg of titanium alloys which was 0.5% of the structural weight, 727 had 290 kg (1%) and 747 had 3850 kg (2.6-2.8%). The more recent Boeing 777 used nearly 50 tonnes of titanium (8.3-10%) and 787 used over 100 tonnes (15%) so it is obvious that the use of titanium alloys in modern commercial aircrafts increased considerably [54,66,67,71]. Titanium alloys are also used in military aircrafts in higher percentage of the structural weight compared with commercial aircrafts in order to withstand larger thermal and mechanical loads associated with extreme manoeuvres and supersonic cruise speeds, because materials and manufacturing costs are less important than performance [67,74]. Structural weight of titanium alloys in modern fighter aircraft is often in the range of 10-50% and its applications include bulkheads, wing torque box, wing spars and fuselage skin [67]. For example, the SR-71 Blackbird had all titanium alloy skin and 93% of its structural weight consisted of titanium alloys. It reached speeds in excess of Mach 3 and temperatures as high as 565°C [67]. Hence, the use of aluminium alloys for supersonic flights is limited. The Boeing

16 F-22 Raptor was the first aircraft with a welded fuselage and 42% of the structural weight consisted of titanium alloys [54,75,76].

2.4 Laser Beam Welding Principles Laser beam welding (LBW) is a high energy density beam (HEDB) fusion welding process which uses a very high intensity laser beam as the heating source. Laser light is monochromatic with a single wavelength, coherent and directional. It can be focused into a very small concentrated spot with very high energy density in the range of 106-107 W/cm2 compared with 103-104 W/cm2 for low energy density arc heat sources, and very low divergence in order to concentrate the laser power at a great distance [77]. It is mostly performed in deep penetration mode by a keyhole mechanism which involves heating, melting and vaporisation of the workpiece. The welding process is significantly influenced by the type of laser used which mainly differ in wavelength and intensity as well as other beam properties such as beam quality (M2) and beam parameter product (BPP). All these parameters influence the focusability and absorptivity of the laser beam. The focusability is an important beam parameter as it controls the spot size. The advantage of operating with a small spot size is that high power density can be maintained even with a considerable work clearance between the focusing lens and the top surface of the workpiece. In other words, the welding process becomes less sensitive to variations in the working distance and avoid spatter from damaging the processing optics [78,79]. However, the use of longer focal length also requires higher laser power to induce keyhole formation for deep penetration welding. The focusability is also referred to as the beam quality, expressed in terms of the beam quality factor (M2), which indicates the extent of deviations of the actual laser beam profile from the ideal shape, so a diffraction limited TEM00 beam with a perfect Gaussian distribution has a value of 1 [16,80]. The beam quality (M2) can be expressed as: 휋 푀2 = 훩휔 Equation (3) 0 휆 where 훩 is the divergence angle, 휔0 is the beam waist and 휆 is the wavelength of the laser. The beam waist and the divergence angle are derived from the power density distribution of the laser beam normal to the beam propagation direction [80]. The beam quality can also be quantified in terms of the beam parameter product (BPP), which determines how small the radius of the beam at its narrowest point known as the beam waist can be focused, for a given divergence angle, with a smaller value indicating higher beam quality. It can be expressed as the product of the divergence angle and the beam waist [16]:

17 퐵푃푃 = 훩휔0 Equation (4)

A high laser beam quality can therefore, be achieved when both the beam divergence and the minimum beam waist diameter are small. The consequence of a high laser beam quality is that the laser beam can be focused on a small point to enable fibre optic delivery of the beam and generate very high powers which lead to deeper weld penetration and narrower weld width with a large depth to width ratio [78]. As only a very limited volume of material is affected by the laser beam due to its focal spot diameter, higher welding speeds can be used and the resulting heat input is small. In addition, the lower heat input results in less thermal stress and distortions, and smaller loss of mechanical properties in the weld metal and the heat affected zone [81]. In contrast, the small focal spot diameter imposes stricter requirements on precise part fit-up and alignment before welding [82].

2.4.1 Conduction Mode and Keyhole Mode There are two different modes of laser beam welding depending on the power density of the focused laser beam on the workpiece. Conduction mode welding occurs when the power density is less than around 105 W/cm2 but also high enough to cause melting of the substrate. Lower power density can be obtained for example, by defocusing the beam to increase the incident laser beam spot size. The power density in this mode is insufficient to cause vaporisation and create a keyhole. The heat from the laser beam is only absorbed at the surface of the workpiece and transferred by conduction into the bulk of the workpiece with a wide heat distribution similar to that in arc welding. The depth to width ratio of a weld produced in conduction mode is typically less than 0.5 with a small penetration depth, and the weld cross-section has a hemispherical weld profile. The weld penetration depth is limited by the thermal conductivity of the material being welded so deep penetration is impossible. Conduction mode welding is a relatively stable process so high quality welds with only few or no welding defects such as hot cracking and porosity can be produced. The process also involves a high heat input which leads to deterioration of mechanical properties, formation of wider FZ and HAZ and greater distortions. As a result, conduction limited laser beam welding is mainly used for spot welding, seam welding and laser cladding [27].

Keyhole mode welding requires a high power density above 106 W/cm2 to form a keyhole in the workpiece. In this mode, the power density is sufficient to cause vaporisation of the metal upon interaction with the laser beam into plume of metal vapour or ionised plasma before significant amount of heat is dissipated by conduction. The vapour pressure from the weld pool pushes out the molten metal and creates a continuous deep narrow cavity called a keyhole, which is filled with metal vapour or plasma and its wall is surrounded by molten metal. It allows the incident laser beam to enter the keyhole and heat not only at the top surface but through thickness of the material so the formation of metal vapour is enhanced. The coupling

18 of the laser beam to the workpiece increases significantly once a keyhole forms, with absorptivity increasing from approximately 3% to 98% [62], due to multiple reflections at the keyhole walls, where the beam energy is reflected repetitively before being able to escape or until all the energy is absorbed by Fresnel absorption [83]. The Fresnel absorption refers to partial reflection and partial absorption of incident light on an opaque surface which are independent of the incident intensity but dependent on the angle of incidence between the laser beam and the keyhole wall, and the wavelength of the laser radiation [83].

For 1 μm wavelength lasers such as Nd:YAG and fibre lasers, the radiation is almost completely absorbed by the Fresnel absorption and plasma effects are smaller because absorptivity is inversely proportional to the square of the wavelength and therefore, the temperature inside vapour plume is not high enough to cause ionisation. Consequently, the threshold value for keyhole formation is lower for shorter wavelength lasers and the process stability is enhanced [84]. For 10 μm wavelength lasers such as CO2 laser, however, ionisation of the metal vapour occurs to form plasma in addition to the Fresnel absorption, which leads to inverse Bremmstrahlung absorption [77]. The inverse Bremmstrahlung is the direct absorption by Joule heating process or re-radiation of a proportion of the laser radiation in the presence of free electrons in the plasma. While it is possible that the energy absorbed by the plasma within the keyhole assists melting of the workpiece by conduction, it can also prevent the full power density of the laser beam from reaching the workpiece and therefore, limit the penetration depth. The keyhole stability depends on the balance of the forces acting outwards from the keyhole and the forces acting inwards from the molten metal around it. The pressure balance in the keyhole can be expressed by Equation 5 [85]:

휌푎푏푙 + 휌푔 = 휌훾 + 휌ℎ푠 + 휌ℎ푑 Equation (5) where 휌푎푏푙 is the ablation pressure, 휌푔 is the pressure due to vapour flow, 휌훾 is the surface tension pressure, 휌ℎ푠 is the hydrostatic pressure and 휌ℎ푑 is the hydrodynamic pressure due to fluid flow around the keyhole. It mainly depends on the balance between the ablation pressure and the surface tension, as these two are the dominant terms [86]. Ablation pressure and vapour pressure prevent the keyhole from collapsing and keep it open. The ablation pressure is caused by vaporisation of molten metal within the keyhole and the vapour pressure is caused by the vapour flow out of the keyhole. In contrast, surface tension, hydrodynamic pressure and hydrostatic pressure tend to close the keyhole cavity. Any changes in the pressure balance may cause keyhole instability and result in keyhole collapse and weld porosity. As the beam is traversed, the molten metal at the leading edge of the weld pool flows around the keyhole to the trailing edge and re-solidifies to form the weld seam which has a chevron mark which points towards the start of the weld [87].

19 2.4.2 Continuous Wave and Pulse Wave Operation Laser beam welding can operate with either continuous wave (CW) or pulsed wave (PW) output laser beam. A PW laser emits short pulses of laser beam at a given power, duration and frequency, so a relatively low total heat input and very fast cooling rate can be achieved to produce a narrow heat affected zone. However, it also has problems related to spatter defect and low depth of penetration. As a result, the PW mode is mainly used for precision welding thin sections due to its ability to produce pulses of high laser powers for very short durations and accurately control the pulse energy. A CW laser produces a continuous stream of laser beam with a constant output power, allowing the formation of a steady molten weld pool during welding, which is more stable than that produced using pulsed lasers. The CW mode is more suitable for high speed welding and has less problems with spatter as well as other welding defects [88].

2.5 Laser Beam Welding Processes There are various types of lasers which can be used for laser beam welding as shown in Table 1. The properties of the laser beam such as the wavelength, beam quality and output power are mainly determined by the active medium used to produce the laser light [77]. The most common industrial lasers for welding are the CO2 and Nd:YAG lasers. The CO2 laser is a type of gas laser which uses a gas mixture of CO2, N2 and He as the active medium to produce the laser light. It is electrically pumped via an electrical discharge within the gas and the beam is amplified by internal reflection within the laser cavity by mirrors placed at either end to form a very concentrated beam of powerful laser light. Pumping is the process of stimulating the dopants to emit photons at a specific wavelength. Since the CO2 laser emits an infrared radiation at a wavelength of 10.6 μm, it cannot be delivered via optical fibre and shows poor transmissivity through glass lenses so special zinc selenide lenses are used instead for the beam delivery [77]. The CO2 laser can produce very high CW power outputs with a decent beam quality in terms of focusability and is more efficient than the Nd:YAG laser. However, almost all metals are highly reflective to this wavelength until a certain power density threshold value is reached and plasma formation at high power density can result in a decrease in weld quality and penetration depth due to loss of laser energy and deterioration of the beam focusing [89].

Table 1 Main laser sources used for high power laser beam welding [27,78] Max. commercially Wall plug Wavelength available output powers M2 efficiency (μm) (kW) (%) CO2 10.60 20 ≥ 1.1 ~10 Nd:YAG 1.06 5 ≥ 35 ~4 Yb-fibre 1.07 50 ≥ 1.1 ~30 Yb:YAG disc 1.03 16 ≥ 6 ~20

20 Nd:YAG laser is a type of solid state laser which produces laser by optically pumping a solid state active medium using flash lamps, arc lamps or diode laser. The laser diodes are commonly used for pumping Nd:YAG laser because of its high beam quality and long lifetime, and to avoid thermal lensing in the active medium which induces a refractive index gradient and modifies the laser beam output [77]. As its name suggests, the active medium is the neodymium ions, Nd3+ embedded within the host lattice, which is an insulating crystal of yttrium aluminium garnet (YAG) with the chemical composition Y3Al5O12 or glass. The host material does not participate in the lasing action but only the dopant does. The active medium is in the form of a rod for the traditional Nd:YAG laser and a thin disc for disc laser, developed to avoid the thermal lensing effect at high power levels by keeping the resulting thermal gradients parallel to the beam propagation direction [77]. The wavelength of the Nd:YAG and Yb:YAG disc lasers are 1.06 μm and 1.03 μm, respectively, which is around an order of magnitude smaller than that of the CO2 laser, and is absorbed better by most metals. Hence, fibre optic beam delivery can be used instead of mirror systems and eliminate the need for mirror alignment. The peak temperatures reached when welding with these shorter wavelength lasers are not high enough to cause ionisation and therefore, forms a plume of metal vapour rather than a plasma so the plasma effects are negligible [89]. The main disadvantages of Nd:YAG laser are the limited maximum power of around 5 kW, much lower beam quality compared with the CO2 laser and low wall plug efficiency, meaning that a significant proportion of the laser power is converted to heat [90]. The reasons for poor beam quality of high power Nd:YAG include the presence of large numbers of higher order modes, thermally induced depolarisation and imperfections in the active medium. The diode laser mentioned above can also be used to produce relatively high laser powers up to around 6 kW but only with low power densities up to 105 W/cm2, meaning a large spot size even at focus, so the welding process is limited to conduction mode welding of thin sheets [91,92]. It uses a semiconductor single emitter diodes as the active medium which is electrically pumped and the beam quality is poor at high output powers due to incoherent superposition of many single emitters.

Fibre laser is another type of solid state laser which uses a silica glass optical fibre as the active medium with a core doped with low levels of rare earth elements such as erbium (Er), neodymium (Nd), thulium (Tm) and ytterbium (Yb), as well as for fibre optic beam delivery [90]. The wavelength of the fibre laser light depends on the rare earth element used as the dopant. Currently available high power fibre lasers mainly use ytterbium because other types of fibre lasers have lower wall plug efficiencies. Ytterbium laser has a wavelength of 1.07 μm which is similar to that of the Nd:YAG laser, so it has the same advantages of shorter wavelength lasers such as fibre optic beam delivery, improved weldability and lower minimum power density

21 threshold for keyhole mode welding compared with the CO2 laser [93]. Pumping is done using a high power diode bars as used in diode pumped Nd:YAG laser or semiconductor single emitter diodes which are coupled into the doped active fibre [90]. The advantages of using the diode laser over diode bars include higher output power, higher efficiency of pumping diodes, no need for alignment optics since all fibre or water cooling used in other diode pumped lasers due to the absence of an optical cavity so lower cooling requirements, significantly greater lifetime over 100,000 hours, reduced maintenance and downtime, and diode replacement costs [94].

The fibre core can either be in single mode or multi mode. In a single mode fibre lasers, the pump light is coupled directly into the core [95], whereas, in a multi-mode fibre laser, the pump light is coupled to a low refractive index external cladding made of undoped glass around the core to restrict the pump light within the double clad fibre and minimise attenuation of this energy [77]. A low power fibre laser, in the order of a few watts, was initially used in the early 1960s for low power applications such as optical signal amplifiers in telecommunications. The higher power 100 W fibre laser only became available much later in 2000 and in recent years, the output powers in the multi kW range up to almost 100 kW, far exceeding those using commercially available Nd:YAG and CO2 lasers have been developed for materials processing through power scaling of the fibre laser [16,96]. The single mode fibre laser has relatively low output powers from a few watts to over 1 kW but improved beam quality over entire power range compared with most commercially available lasers, around 10 times better than the

2 Nd:YAG laser and even exceeding those of the CO2 laser, with a typical M less than 1.07, close to a pure Gaussian distribution. A very fine focus spot size are produced using the fibre laser, to about half the size of the other conventional lasers due to a symmetric beam on either of the axes, small active fibre core diameter and shorter wavelength [16]. As a result, much higher power density can be obtained with the same power, which contributes to deeper penetration and faster welding speed of around 50-100% more than the Nd:YAG laser, and thus reducing the overall heat input [97]. In addition, a smaller 0.3 mm diameter beam delivery fibre can be used compared with 1 mm for the Nd:YAG laser [90].

Much higher output powers suitable for deep penetration keyhole mode welding of metals are produced by splicing the output from a series of single mode fibre laser modules in parallel using a beam combiner into one multi mode fibre laser [94,98]. The individual single mode modules or the pump diodes can easily be repaired and replaced since they are modulised. Although the power scaling technique reduces the beam quality as the beam is no longer single mode, the reduction is relatively small and the resulting beam quality with M2 of 7-10 is better than high power Nd:YAG laser as shown in Table 1.

22 Fibre laser has received increasing interest for laser beam welding process based on the advantages it offers over existing lasers such as better laser efficiency, higher output power, superior beam quality, longer working distances, lower maintenance, reduced cooling requirements, smaller footprint and more compact design [78,79,90]. As the entire lasing fibre is used to produce the laser light, the heat generated in the fibre laser dissipates effectively over the entire length of the long and thin fibre so the cooling requirements and thermal effects due to pumping are minimised and the wall plug efficiency is considerably greater than the conventional lasers [95]. The maximum power density of the multi-mode fibre laser is almost as high as that of electron beams with similar aspect ratios, so they provide the advantages of lowering the capital investment, operating without a vacuum and increasing the productivity [99]. However, since the fibre laser is a relatively new technology, it is not widely used yet in the aircraft industry as the initial investment costs are high and its benefits need to be justified before the investment can be made. Not only that, but it is also very time consuming and costly to evaluate the performance of a new technology. In fact, fibre laser welding AA 2024-T3 and Ti-6Al-4V is relatively untested and unproven so there is not much information available. Therefore, investigations still need to be conducted to study the properties and performance of fibre laser welded AA 2024-T3 and Ti-6Al-4V, optimise the welding procedure and processing parameters to consistently produce high quality welds with no welding defects and with good mechanical properties [94].

There has been a significant amount of research conducted on laser welding using fibre lasers in the literature since high power single-mode fibre lasers became available and around 70 publications related to the subject have been found. 40 of those were related to welding steel, 4 on magnesium alloys, 2 on nickel base superalloys, 17 on aluminium alloys and 8 on titanium alloys. Only one publication was found on fibre laser welding AA 2024-T3 and even that was for dissimilar welding AA 2024-T3 and AA 7050-T76. Similarly, only five publications were found on fibre laser welding Ti-6Al-4V, including one for dissimilar welding Ti-6Al-4V and Inconel 718 nickel alloy. Research progress on laser welding AA 2024-T3 and Ti-6Al-4V including fibre lasers is dealt with separately in Section 2.5.1 and 2.7. Miyamoto et al. [15] were one of the first to use fibre laser for welding in 2003. A low power output 40 W CW single mode Yb fibre laser with a M2 less than 1.1 was used at high welding speeds, ranging from 12 to 60 m/min, to weld SUS304 stainless steel foils of 10-100 μm thickness. Since then, higher power output fibre lasers have been widely used to weld a range of mild and low carbon steels such as X65-X100 pipeline steels in conduction and deep penetration welding modes [16,100– 113]; stainless steels such as ANSI 304 welded bead on plate [90,114–126], dual phase and high strength low alloy (HSLA) steels [127–130], galvanised steels [131–133] and other structural steels such as DH36 [134] for applications mainly in the oil and gas industry and

23 shipbuilding industry. Other researchers used the fibre laser for welding thin plates of magnesium alloys for example, AZ31 [135–138], and nickel base superalloys such as Inconel 617 [139,140].

2.5.1 Friction Stir Welding Process Friction stir welding also is an alternative method to mechanical joints. It is an advanced solid state welding process which has the advantage of reducing the occurrence of metallurgical defects such as hot cracks and pores. The process is suitable for non-laser weldable alloys including AA 2024-T3. Unlike laser beam welding, it does not require the use of filler metal or shielding gas. A fine recrystallized microstructure is formed in the weld with minimum component distortion. High static and dynamic properties of the weld structure with a typical joint strength of around 80-100% yield strength of the parent material can be achieved. Unfortunately, a special clamping system is necessary when friction stir welding and also it is only used for simple joint geometries especially butt joints.

2.6 Research Progress on Laser Welding of AA 2024-T3 AA 2024-T3 was not considered weldable by arc welding processes, so it had very limited use for welding applications especially in any stress environment. On the other hand, laser welding of aluminium alloys in general, has been performed for many years as early as 1980s. Huntington and Eagar [141] conducted welding trials on 6.4 mm thick pure aluminium 1100 and 5456 alloy using a CO2 laser at low laser powers of 0.2 and 1.3 kW. They discovered that the joint geometry and surface conditions had significant influence on laser beam absorption.

CO2 lasers have been predominantly used until the late 1990s, when higher power output solid state lasers such as Nd:YAG lasers became available. Leong et al. [142,143] reviewed the welding parameters needed to obtain consistent laser welds in 5182 and 5754 alloys for automotive applications, using both CO2 and Nd:YAG lasers. Zhao et al. [35] published a review of numerous studies conducted on laser welding of automotive aluminium alloys before 2000, including 2008, 2010, 5083, 5086, 5251, 5456, 5754, 6013, 6061, 6063, 6082 and 6111 alloys, using both CO2 laser and Nd:YAG laser. In addition, they also conducted welding experiments using a 3 kW CW Nd:YAG laser on 5182 alloy in conduction mode and measured the vaporisation rate of magnesium in weld metal and compared with the magnesium content in the base metal [42]. Considerable amount of research has been conducted on CO2 and Nd:YAG laser welding of mostly 5000 and 6000 series aluminium alloys at least over the last 15 years due to their good weldability compared with the crack sensitive 2000 series alloys, some of which include laser welding of 5052 [144], 5083 [144–147], 5356 [33,148], 5754 [149] and 5J32 [150] alloys of the 5000 series; and 6056 [33,151,152], 6061 [153] and 6156 [152] alloys of the 6000 series. Few other publications also covered welding of 7000 series aluminium alloys such as CO2 laser welding of 7075 by Liu et al. [154], and CO2 and Nd:YAG

24 laser welding of 7475 by Weston et al. and Squillace et al. [146,148]. More recently, Chen et al. [72] and Cui et al. [155] conducted investigated Nd:YAG laser welding of an aluminium lithium alloy, 5A90.

Far less work on has been done on laser welding AA 2024-T3. Bardin et al. [156] used a 4 kW Nd:YAG laser to weld 13 mm thick AA 2024-T3 in conduction mode so only partial penetration was achieved. Hu and Richardson [157] also attempted autogenous welding 3 mm thick AA 2024 sheets using a 3 kW Nd:YAG laser. Severe keyhole instability was observed at lower heat inputs, which resulted in poor weld seam quality and porosity formation. The welding process was relatively stable when using higher heat inputs and the welds contained less porosity and the keyhole fluctuation was smaller. The factors causing the keyhole instability were determined to be high reflectivity of AA 2024 which effectively raised the power density threshold for keyhole formation, and the presence of low boiling point alloying elements such as magnesium, and therefore, indicated that a process optimisation was needed. A high power diode laser was employed by Sánchez-Amaya et al. [147] to bead on plate weld 2 mm 2024- T3 and five other aluminium alloys in conduction mode again using laser powers from 1.5 to 2.75 kW and welding speeds from 5 to 25 mm/s. It was found that the 2024 alloy had fairly poor weldability in terms of penetration values and solidification cracking sensitivity. Some of the works by the same researchers including Alfieri et al. [26,158] and Caiazzo et al. [159] involved experiments on laser welding 3.2 mm thick AA 2024 using a Yb:YAG disc laser. They discussed the importance of adequate joint preparation to improve weld quality and also examined the effect of heat input and beam defocusing on the weld quality in terms of macro porosity content. It was found that higher heat inputs and defocusing led to lower porosity content due to enhanced convection vortex structure in the weld metal as well as keyhole stability. Tensile strength of welded joints was measured to be greater than 2/3 of the BM and therefore, gave ground for industrial application. Due to the fact that AA 2024-T3 has a very high susceptibility to hot cracking, some researchers such as Jones et al. [160] and Sutton et al. [161] studied friction stir welding (FSW) of AA 2024-T351 as an alternative solution. The problems of formation of brittle solidification products and grain boundary liquation cracking were minimised because friction stir welding is a solid state welding process, and therefore, it was determined to be a viable solution for certain applications. Furthermore, Wang et al. [162] studied variable polarity plasma arc (VPPA) welding of AA 2024-T351 with 2319 filler metal.

2.6.1 Fibre Laser Welding of AA 2024-T3 Even less research has been done on fibre laser welding of AA 2024-T3, where only one publication was available up to 2015 related to this topic. Enz et al. [163] studied fibre laser welding of a dissimilar T-joint made of 2 mm thick AA 2024-T3 and AA 7050-T76, welded with a 1.2 mm diameter 4047 filler metal. A range of laser powers from 3.5 to 8 kW, welding speeds

25 of 2 to 8 m/min and focal positions of 0 to +8 mm, wire feed rates of 2 to 8 m/min and helium shield gas were used. They showed that the single-sided T-joints produced using the high power fibre laser had significantly less macro pores and micro pores than Nd:YAG laser welded AA 6082-AA 5082 dissimilar joints and CO2 laser welded AA 2198-AA 2196 dissimilar joints, whereas, the successively welded double-sided T-joints had a significantly higher porosity content due to poorer degassing condition for double-sided welding. Consequently, fibre laser welding of AA 2024-T3 is not well understood and requires much to be investigated.

Meanwhile, extensive research has been conducted on fibre laser welding of other aluminium alloys. In particular, the 5000 and 6000 series alloys have been studied by many researchers, including AA 6013 [164–171], AA 6061 [172], AA 5052 [173–175], AA 5A06 [176] and AA 5083 [168]. Paleocrassas and Tu [177] investigated low welding speed and low laser power welding of AA 7075 using a 300 W single mode Yb fibre laser for speeds from 1 mm/s to 10 mm/s, which below the minimum speed threshold of 1 mm/s became very unstable due to overheating of the molten pool. Also, Zhang et al. [178,179] studied fibre laser welding of 20 mm thick AA 7075 but using a higher laser power 4 kW fibre laser. Allen et al. [180] used a high power fibre laser for laser welding and hybrid laser-metal inert gas (MIG) welding of 7000 series aluminium alloy of 6 and 12 mm thickness but the exact processing parameters were not stated. Brown et al. [181] used a low power 600 W fibre laser for welding unalloyed AA 1100 and was able to obtain uniform high aspect ratio welds. Shiganov et al. [182] compared welding of a novel Al-Mg-Sc-Zr alloy, the AA 1570 alloy using a 3.5 kW fibre laser and a CO2 laser for aircraft applications. The use of the fibre laser for welding 2 mm thick AA

1570 increased efficiency by 25-30% in comparison with the CO2 laser, where the power density required for starting penetration using the fibre laser was around 50% lower than using the CO2 laser.

2.7 Research Progress on Laser Welding of Ti-6Al-4V

Previous studies on welding Ti-6Al-4V mainly utilised CO2 lasers until between the late 1990s and the early 2000s, when high power solid state lasers such as Nd:YAG suitable for welding, became available. One of the first attempts at welding Ti-6Al-4V was made by Li et al. [183] who conducted experiments using a 2.5 kW CW CO2 laser and a 1 kW PW Nd:YAG laser to weld 1.6 and 2.0 mm thick sheets. It was found that welds without undercut and with acceptable porosity were obtained using the CO2 laser when welded with a filler metal and - 1 mm negative defocus below the top surface of the workpiece. Similar experiments were conducted on commercially pure titanium plates of 0.5 mm and 1.0 mm by Yunlian et al. [184] using a 2 kW CO2 laser and Liu et al. [62] using Nd:YAG laser of various output energy from 180-300 A. Joint strengths similar to that of the base metal were achieved but the process was found to be unstable and therefore, required further investigations. Wang et al. [185]

26 studied the influence of increasing temperatures up to 450°C on the tensile properties of CO2 laser welded 3.3 mm thick bead on plate welds produced using a laser power of 2.5 kW, welding speed of 1.5 m/min and helium shielding gas. The tensile strength of the welds was greater than that of the BM for all temperatures due to the harder martensitic microstructure in the FZ. However, at above 300°C, the yield strength of the welds decreased due to beta precipitation in the FZ. Caiazzo et al. [186] and Casalino et al. [187] studied the influence of various welding parameters including laser power, welding speed and type of shielding gases on CO2 laser welding Ti-6Al-4V sheets of different thickness ranging from 1.0 to 2.85 mm. In general, lower levels of welding imperfections were found when using smaller sheet thickness, higher welding speeds and helium instead of argon as the shielding gas due to reduced plasma effect. Chen et al. [188] also performed similar experiments on BT20 titanium alloy using 3 kW CO2 laser and confirmed that the joint quality can be controlled by using a suitable heat input as a function of laser power and welding speed.

Most of the publications from 2007 until the early 2010s reported on welding Ti-6Al-4V using the newly developed Nd:YAG lasers instead of the CO2 lasers. Casavola et al. [189] examined the static and fatigue behaviour of 3 mm thick butt welded Ti-6Al-4V with CW Nd:YAG laser using a laser power of 2 kW and a welding speed of 1.5 m/min and argon shielding gas. Their results showed that fatigue failures occurred in the BM and not in the weld metal. However, the welding parameters were not optimised and so, Ti-6Al-4V was found to be more difficult to weld than CP Ti. Cao et al. [190] and Kabir et al. [191] investigated the influence of welding speed and defocusing distance on microstructure and properties of Ti-6Al-4V sheets using 4 kW Nd:YAG laser and proved that the Nd:YAG laser is an applicable laser source for welding Ti-6Al-4V, but the weld quality was strongly dependent on the welding speed. As a result, a significant loss of ductility was observed in the welded joints due to the presence of micro pores, underfill and oxide inclusions. It was found that the porosity area decreased while hardness increased with increasing welding speed. Serroni et al. [192] conducted experiments on Nd:YAG laser welding 2.5 mm Ti-6Al-4V using 1.2-1.7 kW laser powers and 10-32 mm/s welding speeds. It was found that the penetration depth was affected by the laser power where full penetration was only achieved above 1.7 kW. Squillace et al. [193] also studied the influence of welding speed and laser power on weld quality of autogenous butt welded 1.6 mm thick Ti-6Al-4V sheets using 2kW CW Nd:YAG laser with 0.8-1.2 kW laser powers and 17- 58 mm/s welding speeds. The weld quality was evaluated in terms of specific heat input and discovered that at higher heat inputs conduction mode was obtained, at lower heat inputs keyhole mode and at intermediate heat inputs, underfill was the largest. Zhao et al. [194] also conducted similar investigations on dissimilar welding of 2mm thick Ti-6Al-4V and lead metal plates using CW 1 kW Nd:YAG laser at low laser powers of 200-250 W and low welding

27 speeds of 20-25 mm/s to examine the effect of heat input on microstructural and mechanical properties.

The effect of joint preparation whose edges were obtained by fibre laser cutting or milling, on the mechanical behaviour of Nd:YAG laser welded 1 mm thick Ti-6Al-4V was studied by Scintilla et al. [195] and comparable joint tensile properties were obtained using fibre laser cutting. A technique for producing high quality welds by modulating the output of Nd:YAG laser beam was evaluated by Blackburn et al. [196] who were able to produce low levels of porosity and undercut with appropriate choice of modulating parameters and defocused laser beam. Other researcher [197–199] investigated PW Nd:YAG laser welding thin sheets of Ti-6Al-4V and found that the ratio between pulse energy and pulse duration were the most important welding parameters in defining the penetration depth, where insufficient laser power and pulse duration led to defects such as lack of fusion.

Further research on CO2 laser welding Ti-6Al-4V were conducted in the early to mid-2010s. Xu [200] investigated the microstructure and mechanical properties of 6 mm Ti-6Al-4V welded using 6 kW CO2 laser and observed slip deformation due to welding residual stress. Jianxun et al. [201] investigated the plastic damage behaviour of CO2 laser welded 2.5 mm thick Ti- 6Al-4V by measuring the microvoid density in different regions of the welded joint using a laser power of 2.5 kW and welding speeds of 1.2-1.5 m/min. The plastic damage along the welded joint from weld metal, heat affected zone and base metal showed a non-linear distribution and the largest damage was observed in the weld metal, which involved microvoid nucleation, growth and coalescence. Balasubramanian et al. [202] compared the performance of 5.4 mm thick Ti-6Al-4V welded joints produced by gas tungsten arc welding, electron beam welding and CO2 laser welding with optimised parameters by trial and error to obtain defect free full penetration welds. The joint strengths of the laser welds were lower than those produced by electron beam welding but greater than those produced by gas tungsten arc welding due to the presence of fine acicular martensitic structure in the weld metal. Brandizzi et al. [203] gave a comparison between the mechanical properties of 3 mm thick Ti-6Al-4V butt joints produced by 6 kW CO2 laser and by laser arc hybrid welding and determined from static draw testing that within the range of parameters investigated, only small differences in the strength and ductility of the joints were observed between the two techniques adopted.

2.7.1 Fibre Laser Welding of Ti-6Al-4V To date, very little has been reported on the weldability of Ti–6Al–4V alloy using high power fibre lasers and so the process is far from being optimised. Mueller et al. [204] used Yb fibre laser to butt weld 1.6 and 6.4 mm thick Ti-6Al-4V using laser powers of 1.3 and 5 kW, respectively. Class A welds specified in the standard AWS D17.1:2001 were only produced using the slowest welding speeds for the 1.6 mm material which had the widest weld width,

28 whereas, higher welds speeds could be used to obtain Class A welds for the 6.4 mm material. However, increasing the welding speed led to a change in weld shape from an hour glass shape to a more nail head shape which was more prone to root porosity formation. They also conducted further experiments using 3.2 mm thick Ti-6Al-4V using 2.5 kW laser power and Class A welds were produced at a welding speed of 3.75 m/min. Costa el al. [205] conducted a research on the weldability of 6.5 mm thick Ti-6Al-4V using fibre laser at laser powers of 4 and 8 kW and welding speeds of 1-3 m/min without defocus and with argon shielding gas. Full penetration welds without defects were achieved using 8 kW and welding speeds above 1.5- 1.8 m/min, whereas, only partial penetration welds at 4 kW even at very low welding speeds. In addition, welding at high power levels and low speeds led to full penetration but with undercuts, excessive penetration and bead decay. Fan et al. [206] studied laser welding of very thin 0.7 mm Ti-6Al-4V using diode laser, PW Nd:YAG laser and 2 kW CW fibre laser at a low laser power of 600 W and a welding speed of 4 m/min and argon shielding gas, which was the only set of welding parameters was used. High levels of porosity were observed in the FZ and they predicted that increasing the welding speed would also increase the probability of pores being retained in the FZ due to insufficient time for gas bubbles to escape the melt pool. Campanelli et al. [207] investigated the effect of welding speed on microstructure and mechanical properties of 2mm thick butt welded Ti-6Al-4V sheets using fibre laser at a constant laser power of 1.2 kW, various welding speeds between 1.0 and 2.5 m/min, -2 mm defocus, argon and helium at the upper and bottom surfaces, respectively at a flow rate of 15 l/min for both. Since only the welding speed was varied, the weld shape was studied in terms of the linear energy. Higher linear energy promoted an hourglass shaped weld cross- section, whereas, lower linear energy promoted a V shaped bead. The tensile strength and ductility were both found to be less than those of the BM. Other weakly related studies involved, for example, dissimilar welding of 2 mm thick Ti-6Al-4V to Inconel 718 nickel alloy using a 1 kW CW fibre laser [208], fibre laser-gas metal arc (GMA) hybrid welding of 1.5 mm CP Ti using a 2 kW CW fibre laser [209], 10 kW fibre laser welding 4 mm thick CP Ti [210] and also 6 mm thick Ti-VT23 [211]. As it can be seen, no comprehensive study on fibre laser welding Ti-6Al-4V and optimisation of the process has been conducted yet.

2.8 Conclusions This chapter reviewed the structure and properties of AA 2024-T3 and Ti-6Al-4V such as their weldability, advantages and disadvantages, and their applications in the aircraft industry. The principles of laser welding were examined including different welding modes, properties of three types of laser sources appropriate for high power laser beam welding and the mechanisms of fibre lasers in detail. The relevant literature regarding the laser beam welding of these materials were reviewed and the progress made in implementing the fibre laser for

29 welding aluminium and titanium alloys was also reviewed. It was evident that very little progress has been made on fibre laser welding AA 2024-T3 and Ti-6Al-4V and therefore, the importance of conducting this research was discussed.

30 3 EFFECT OF WELDING ON MICROSTRUCTURE OF AA 2024-T3 AND TI-6AL-4V

3.1 Introduction The quality and performance of welded joints depend on the weld geometry, melt pool behaviour during welding, the metallurgy of the weld and the heat affected zone, and welding defects. The complexity of chemical and metallurgical actions which take place during welding may result in subsequent failure of the weld in service and so, it is important to anticipate and incorporate the effects of welding at the design stage. According to the American Society of Mechanical Engineers (ASME), 45% of the causes for welding imperfections is due to poor process conditions [212]. In fact, because the research on the development of welding techniques has been largely conducted to satisfy the needs of the industry for demonstrating the maximum capabilities of a process, many welding fundamentals have not been researched and therefore are not yet fully understood [31]. This means that a significant amount of work is required to be done to be able to predict and optimise the laser welding process to produce consistently quality welds.

In order to obtain an acceptable weld profile and satisfactory mechanical properties, control of weld bead shape is essential as the mechanical properties of welds are affected by the weld bead shape. The weld bead shape which affects the weld metal solidification behaviour is influenced by welding parameters and the corresponding amount of heat input into the workpiece [213]. Therefore, it is necessary to determine the influence of the welding parameters including laser power, welding speed, defocusing distance and shielding gas [144,214] on weld morphology as well as to identify the sources of welding defects. It would then be possible to identify the optimum combination of welding parameters which ensures the required weld quality and properties, and also minimises welding defects [72,188,215].

Schuöcker [82] classified the welding parameters into three categories: design parameters, technological parameters and metallurgical parameters. The design parameters describe the geometry of the workpieces in the vicinity of the weld such as gap width and joint geometry. The gap width for example, must be smaller than the focused spot size and should be between 0.05-0.15 mm for a good weld seam quality [216]. The technological parameters cover all properties of the laser beam, shielding gas and filler metal supply and other related factors. These mainly include power density, laser power, mode, wavelength, focal position, focal length and welding speed. Defocusing the beam can produce a wider melt pool and larger keyhole which may improve the process stability. Welding speed should be high to reduce welding costs per unit length but also within the optimum range to produce the required weld

31 quality. The risk of cracks and porosity depend largely on the welding speed. The metallurgical parameters are related to the material properties including chemical properties, thermal and mechanical properties. According to Schuöcker [82], the weld seam geometry is affected significantly by all the above parameters, those related to laser beam characteristics, the process parameters, material characteristics and the joint design.

This investigation was concerned with the welding parameters including power density, laser power, welding speed, focal distance, welding with or without filler wire, and type of shielding gas; and their effects on the weld shape and the final solidification structures of AA 2024-T3 and Ti-6Al-4V welds which further influence the overall mechanical properties of the welds. Integrated parametric studies which focus on the influence of various welding parameters on the welding results for fibre laser welding of AA 2024-T3 and Ti-6Al-4V have been rarely published. Most often, only a small number of case-based parameter variations has been investigated partly because high power fibre laser is a relatively new technology and as the initial investment cost is high, only a few existing laser welding systems have been replaced by fibre laser. However, even for other types of lasers, they have often been investigated case- specific with a certain range of laser power, focal position and welding speed, which makes it difficult to transfer the results.

Previous studies concerning the effect of welding parameters on weld seam geometry have already determined different shapes of laser welds in terms of the top and bottom weld widths. The work of Karlsson et al. [104–106,109,217] for example, investigated the influence of laser welding parameters on the weld seam geometry of high strength steels welded with 15 kW fibre laser and it was found that increasing the welding speed can suppress root sagging and undercuts while decreasing the welding speed can suppress lack of penetration. Manonmani et al. [218] studied the effect of laser power, welding speed and beam incidence angle on the weld seam geometry in terms of the penetration depth, the bead width and the area of penetration of a 2.5 mm thick AIS1304 stainless steel. They showed that the depth of penetration and penetration area increased with the laser power and the beam angle, whereas, the weld width decreased with increasing welding speed. A trend was observed where the penetration depth and area increased to a maximum value and then decreased with any further increase in welding speed. This was due to the fact that the mode of heat transfer changes from a keyhole mode at lower speeds to a conduction mode at higher speeds.

The convective heat transfer and fluid flow in the weld pool are controlled by the welding parameters which in turn influence penetration depth, shape and solidification structure of the welds [219]. The effect of welding speed on the weld shape and solidification structure of the fusion zone in aluminium alloy 6060 was observed by Coniglio et al. [220–222] who demonstrated that increasing the welding speed modifies the weld pool shape from elliptical

32 to teardrop, in which the columnar grains are essentially straight in order to grow normal to the pool boundary with transverse strains and as a result, increases cracking susceptibility. The cause of solidification cracking can be attributed to insufficient weld bead size or inappropriate weld shape, where the weld bead is unable to withstand the contraction stress during solidification because the molten metal in the weld pool cannot fill in the gaps due to inadequate supply or narrow channels between solidifying grains. AA 2024-T3 in particular, has a low melting point and ionisation energy, and a high thermal conductivity so it is very sensitive to variations in heat input during welding. It also contains a small amount of magnesium which increases crack sensitivity by broadening the solidification range that induces hot cracking. Silicon on the other hand, at a concentration greater than 4% reduces the solidification range and improves the crack sensitivity, and this level of silicon content in the weld can be obtained by welding with Al-Si filler metal. In this way, the chemical composition of the weld can be adjusted beyond the crack sensitive range and compensate for the loss of alloy elements by vaporisation. Similarly, Dixon et al. [223] also attributed the cause of solidification cracking in laser welded thick section steel to the weld shape. The high depth to width ratio of laser welds causes restraint and induces highly localised thermal strain across the welded joint which leads to centreline cracking. Fortunately, full penetration laser welds unlike partial penetration welds, are less prone to solidification cracking because the high restraint at the weld root can be avoided. Therefore, the weld shape must be carefully controlled to prevent solidification cracking and also other welding defects.

Vitek et al.[224], for example, developed a statistical model using neutron network (NN) to predict the weld pool shape parameters (penetration, width, width at half-penetration and cross-section area) in pulsed wave Nd:YAG laser welds of aluminium alloy 5754 by considering the welding speed, average power, pulse energy and pulse duration. Lee et al. [225] used the Taguchi method and regression analysis to optimise Nd:YAG laser welding of commercially pure (CP) titanium. It was found that the laser pulse width and focal position had the greatest effect on the control of weld seam length to produce a sound weld quality.

Different top and root shape classes of laser welded Ti-6Al-4V have been identified in published literature. The centre of the fusion zone as observed by Balasubramaninan et al.

[202] in CO2 laser welded Ti-6Al-4V joints, generally presented a convex shape at middle thickness due to volume contraction, surface tension and phase transformation during welding. The specific heat input as mentioned above has a strong impact on the welding bead shape. An increase in welding speed, due to the lower value of specific heat input transmitted to the workpiece, leads to a reduction of the weld width. The heat input was found to be a highly influencing parameter for the bead shape by Squillace et al., [193] who studied the effect of welding parameters on morphology and mechanical properties of fibre laser welded Ti-6Al-4V

33 butt joints. Higher heat input promoted an hour glass shaped weld bead whereas, lower heat input promoted the formation of a V-shaped weld bead. Campanelli et al. investigated fibre laser welding of Ti-6Al-4V at a constant power of 1.2 kW and they observed a change in the weld shape from nail head to V-shape when the welding speed was increased from 2 to

2.5 m/min (22). Mueller et al. [204,226] studied the potential application of CO2 laser and fibre laser welding of Ti-6Al-4V for aircraft applications and observed a trend where a change in the weld shape from an hourglass shape at low speed to nail head shape at fast speed increase the tendency to entrap gases and form root porosity. The same trend was also observed by

Chen et al. [72] in their investigation on CO2 welding BT20 titanium alloy. Hilton et al. investigated fibre laser welding of 3 and 5 mm thick Ti-6Al-4V and linked the weld profile to porosity level. Low levels of porosity were found in the 5 mm thick weld which used high laser powers. Interestingly, the measured face weld width was smaller than the root weld width due to the keyhole behaviour, which caused a larger molten volume in the lower part of the weld and promoted the escape of any trapped root shielding gas [227].

The investigation presented in this chapter was therefore conducted to determine the influence of key fibre laser welding parameters on the resultant weld quality when welding AA 2024-T3 and Ti-6Al-4V, the formation of welding defects and to develop techniques to avoid these issues.

3.2 Materials and Experimental Procedures

3.2.1 Materials Heat treatable aluminium alloy (Al-Cu-Mg) alloy 2024 sheets in the T3 temper condition (i.e. solution heat-treated, cold worked and naturally aged) with 3 mm thickness and mill-annealed titanium alloy Ti-6Al-4V (Grade 5) sheets with 2 mm thickness were used. The nominal composition of each alloy is given in Table 2 and Table 3. A 1.0 mm diameter consumable 4043 aluminium filler alloy of a nominal composition Al-5%Si was used when welding 2024 with a filler wire. All Ti-6Al-4V welding trials were autogenous.

Table 2 Chemical composition of AA 2024-T3 (Wt. %)

Al Cu Mg Mn Cr Fe Si Ti Zn Balance 3.8-4.9 1.2-1.8 0.3-0.9 0.1 0.5 0.5 0.15 0.25

Table 3 Chemical composition of Ti-6Al-4V Grade 5 (Wt. %)

Ti Al V Fe O Balance 5.5-6.76 3.5-4.5 0.25 0.2

34 3.2.2 Fibre Laser Beam Welding Process A 5 kW continuous wave (CW) ytterbium fibre laser system YLS-5000 from IPG Photonics was used in TEM01* mode for laser welding. The beam output of 200 W single-mode active fibres was combined to deliver the beam output to the workpiece via a feed fibre with a diameter of 200 µm. To protect the feed fibre from damage during welding, the feed fibre was coupled into a processing fibre with a diameter of 300 µm which was connected directly to the laser processing head [138]. According to Kratky et al. [138], although the input beam diameter from the feed fibre is smaller than the core diameter of the processing fibre, the beam divergence remains unchanged whereas the output beam diameter increase to that of the core diameter. The beam diameter at focus was 630 μm. The wavelength of fibre laser was 1070 nm, the beam quality factor (M2) was around 7.3, the divergence half angle of the focused beam was 12.5 mrad and the Rayleigh length for the multimode beam, scaled with M2 as illustrated in Equation 3, was around 22.6 mm. The focal length of focusing lens was 300 mm and the diameter of the focusing lens was 50 mm. The focal length of the collimator lens was 100 mm and the diameter of the collimator lens was 50 mm. A beam parameter product (BPP) of less than 2.5 mm·mrad was formed.

Most lasers operate in either TEM00 or TEM01* mode, which represents the transverse electromagnetic mode pattern measured in a plane perpendicular to the propagation direction of the laser beam. TEM00 mode is the fundamental transverse mode of the laser resonator with a Gaussian distribution. TEM01* mode, also known as doughnut mode, is made from an oscillation between two orthogonal TEM01 modes, where 0 is the number of radial zero fields and 1 is the number of angular zero fields [228].

TEM00 TEM01 TEM01* Figure 3 Intensity distributions for various beam modes

Figure 3 shows the intensity distributions and patterns for various beam including the TEM01* mode used in this investigation. The radius of the beam area which contains 86% of the beam energy is larger by 1.32 times for the TEM01* mode than for the TEM00 mode [229]. In most lasers, higher modes mean more power and also multi-mode fibre lasers with a top hat beam profile are not susceptible to spiking discontinuities observed in the Gaussian beam profile single-mode fibre laser welds [230]. The beam quality factor, denoted M2 which determines

35 the focusability of a beam, is equal to one for a Gaussian beam and 1.05-1.3 for a single-mode fibre laser. The M2 value of a multi-mode laser according to Quintino et al. [16] is reduced to around 7-10 as a result of combining single-mode fibre laser outputs into a multi-mode fibre. Still, the resulting multi-mode fibre laser beam has small beam divergence and focal spot diameters suitable for deep penetration welding and so the use of the TEM01* mode was preferred for high power laser welding. a) Top and side shielding gas supply

Fibre laser supply

Filler wire Supply

Clamping Root Workpiece bars shielding gas supply b)

Figure 4 a) Experimental setup and b) a cross-sectional view of the fixture used for automated fibre laser welding of AA 2024-T3 and Ti-6Al-4V thin sheets

The laser processing head was mounted on a 6-axis articulated KUKA robot. The workpiece was clamped using steel top clamping bars to a heavy-section steel backing plate with a copper insert in which a 10 mm wide and 10 mm deep slot was machined to supply root shielding gas as shown in Figure 4. A BINZEL Master-Feeder system was used to supply filler metal into the leading edge of the weld pool, ahead of the laser beam impingement point at an angle of 45 or 60° to the workpiece. The processing head was tilted at 5° to avoid laser beam back reflections from damaging the optics and also re-entering the laser cavity and affecting the beam [81]. An air-knife was located directly behind the focusing lens which blows perpendicular to the optical axis of the laser beam in order to protect the lens system from

36 spatters. The air-knife was far enough from the workpiece so that it has no influence on the welding process.

Industrial grade argon and helium with 99.999% purity were used for welding AA 2024-T3 while only argon was used for welding Ti-6Al-4V. The shielding gas was supplied to protect both top and underside of the weld. The coaxial shielding gas was delivered via the weld nozzle to protect molten pool and the back protecting shield gas was supplied via the shielding gas path in the copper insert to protect back weld. Since Ti-6Al-4V is highly reactive with oxygen and forms brittle intermetallic phases at high temperatures, a total shielding of titanium weld seam was provided by the additional trailing shielding shoe with its own shielding gas supply following along behind the laser beam at a distance of no more than 2 mm from the surface of the workpiece. The effectiveness of gas protection in titanium welds was determined by observing the colour of the weld surface where silver is ideal, yellow is acceptable and white or blue are poor [231]. Argon was used for back shielding in all experiments at a flow rate of 15-20 l/min. The flow rate for coaxial or side-jet shielding was 20-25 l/min and 25 l/min for drag cover shielding. The gas flow rates were varied depending on welding parameters used. The shielding gas lines especially the underside shielding were sufficiently purged before welding.

Prior to welding, the specimens were visually inspected for any surface contaminations or misalignments. Contaminations such as dust, oil and surface oxides are a major source of weld porosity and therefore, it was important to remove these from the welding surfaces. According to Cao et al. the surface condition may also influence the energy absorption of incident laser beams and the threshold power density for keyhole welding [232]. For AA 2024- T3, both top and bottom surfaces of each specimen were brushed with a stainless steel wire brush and then cleaned using an industrial grade unbuffered 99.9% pure acetone before welding. On the other hand, for Ti-6Al-4V, physical surface treatments have been reported to increase the risk of weld porosity [188,215] and also, any subsequent increase in surface roughness may increase adsorption of moisture from the atmosphere on the surface of the specimens. While Ti-6Al-4V has a good resistance to oxidisation, hydrogenation and nitrogenation at room temperature, the absorption of oxygen, hydrogen and nitrogen increases with temperature. It begins to absorb hydrogen at 250°C, oxygen at 400°C, and nitrogen at 600°C and interacts with oxygen more intensively within the range of 800-900°C [215]. Oxygen can lead to embrittlement of Ti-6Al-4V welds and nitrogen can increase tensile strength and hardness of the weld, but also reduce ductility with a more serious effect on embrittlement.

The absorption of hydrogen and the formation of low strength acicular titanium hydride, TiH2 are also a source of hydrogen embrittlement. In addition, the welds absorbing these gases are prone to porosity. Instead, based on the work by Chen et al. [188], a five step chemical method

37 of degreasing, rinsing, pickling, rinsing and drying procedures were used to prepare the Ti- 6Al-4V specimens. To clean the surface free of oil and dust, an alkali solution of 5-10% NaOH in acetone or alcohol was found effective. To remove the oxides, pickling with an acid solution of 10% HF, 30% HNO3 and 60% distilled water lasting for 2-5 minutes was used. Chen et al.

[72] observed a reduction in weld metal porosity by the removal of TiO2 layers, which are hygroscopic and adsorb moisture from the atmosphere when CO2 laser welding titanium alloy. Rinsing was done using acetone and welding was performed within 72 hours from drying.

Welding was performed parallel to the rolling direction of the AA 2024-T3 and Ti-6Al-4V sheets, and followed the recommendations written in the standards ISO/TR 17671-6, AWS C7.2 and BS EN 1011-6 for the control of laser beam welding of metallic materials [233–235]. All the specimens which were used for microstructural analysis were welded bead on plate so there was no requirement for joint fit up or alignment tolerances during weld preparations. The main welding parameters investigated were power density, laser power, welding speed, focal distance, filler metal feed rate, joint geometries, welding with or without filler metal and type of shielding gas. All Ti-6Al-4V specimens were autogenously welded using argon shielding gas only.

The laser power was directly obtained from the laser power indicator of the fibre laser welding system and the welding speed was controlled by the robot. The focal distance and focal spot diameter were accurately determined using PROMETEC’s laser scope UFF100 type beam mass analyser and PROLAS software.

3.2.3 Metallographic Specimen Preparation Preparation of weld specimens for metallographic examinations was conducted by following the methods specified in the standards, ASTM E3 [236] and BS EN 1321 [237]. Two transverse cross-sections were taken from the welds by electrical discharge machining (EDM) for all welding parameters set, one for examining the weld seam top surface and the other for examining the weld cross-section. Metallographic specimens were mounted using hot mounting technique where the sample was placed in a mounting press with Bakelite resin and cured under heat at 180°C and pressure of 250 bar for 5.5 minutes and cooled for 2 minutes. The mounted samples were then ground with abrasive silicon carbide, SiC, papers in successively finer steps up to 2400 grit to remove material from the specimen surface until co- planar and the quality needed is achieved. After grinding the specimens, polishing was performed with diamonds of 3 μm grain size and lubricants on cotton cloth with moderate pressure on the disc. For final polishing, suspensions of 1 μm grit diamond and lubricants on short fibre velvet cloth were used to produce a scratch free mirror finish. The polished specimens were then chemically etched for optical microscopy. The microstructural constituents of the weld were revealed by using suitable chemical etchants. Keller’s reagent

38 which is a mixture of 95% distilled water, 2.5% HNO3, 1.5% HCl and 1.0% HF, was used to etch AA 2024-T3 weld specimens by immersing for 10-30 seconds fresh. Kroll’s reagent which is a mixture of 92% distilled water, 6% HNO3, 2% HF, was used to etch Ti-6Al-4V weld specimens by dipping for 15 seconds [238]. When it was necessary, etching was performed for a longer duration to correctly reveal the true metallographic microstructure.

3.2.4 Experimental Procedures Experiments were conducted in accordance with the requirements specified in BS EN ISO 15614-11 and BS EN ISO 13919-2 for both Ti-6Al-4V and AA2024-T3 welds welded with or without additional filler wire. These standards provide guidance on levels of imperfections in electron and laser beam welded joints in metallic materials. There are three quality levels given in the standard BS EN ISO 15614-11 identified as Moderate (D), Intermediate (C) and Stringent (B) which refer to the quality of welded joints and not the fitness-for-purpose of a component [239]. The choice of quality level for this investigation was Moderate (D) and the relevant examination and tests for welds in accordance with acceptance level D are listed in Table 4. Radiographic and ultrasonic examinations, and surface crack detection techniques are not mandatory in level D but it would be necessary to also conduct these tests once the welding parameters have been optimised and the final component has been welded. Economic factors are important which includes not only the cost of welding but also that of weld quality inspection.

Table 4 Examination and tests for welds in accordance with acceptance level D specified in BS EN ISO 15614-11:2002 [239]

Test piece Type of examination and test Extent of examination and test Visual examination 100 % Radiographic examination if required Butt weld Ultrasonic examination if required Surface crack detection if required Metallographic examination 1 section minimum Visual examination 100 % Ultrasonic examination if required T-joint Surface crack detection if required Metallographic examination 1 section minimum

As Table 4 shows, visual examination on the welded sheets and metallographic examination on welded specimens prepared in Section 3.2.3 were conducted. Since a minimum of one section extracted from the weld was required for metallographic examination, it was necessary to check the conformity of the weld seams. A single weld cross-section per welding parameter set could not guarantee the uniformity of weld seam geometry over the whole weld seam. In order to obtain representative trends in variations of the weld seam geometry for varying welding parameters, the conformity of weld seam geometry had to be checked. The possible methods of checking the conformity of weld seam geometry over the whole weld include

39 extracting sections parallel to the welding direction or making a large number of samples with the same parameters. However, since a large number of parameter variations were investigated in this study, it was not practical to follow these methods which consume a lot of time, effort and material costs. Instead, the shape and the quality of the top and bottom weld seam surfaces were used as an indicator for the conformity of the weld seam geometry. Random inspections of weld seams showed a positive relationship between the quality of the top and bottom weld seam surfaces and the uniformity of the weld geometry. Three cross- sections were extracted; one at the centre, one near the weld start and the other close to the weld end positions. The shape of the weld seams remained fairly constant and only small variations were observed on the top and bottom weld seam surfaces as shown in Figure 5. On the other hand, for welding parameter sets which produced poor weld seam quality, the source of weld imperfections was readily identifiable from the top and bottom weld seam surface due to the fact that the welded sheets were thin and in most cases fully penetrated as also shown in Figure 5. Therefore, the shape of the top and bottom weld seam surfaces were used to evaluate the conformity of the whole weld, making it possible to assume that the chosen weld cross-section is representative for the whole weld.

a) b)

Figure 5 Conformity of weld seam geometry represented by a) a good weld surface quality and b) a poor weld surface quality.

Macroscopic and microscopic inspection of transverse sections of metallographic welded specimens followed the test procedures specified in BS EN 1321 [237]. Macroscopic examination was conducted under low magnification with etching and microscopic examination with a magnification within 50 to 500 times also with etching to reveal features of welded joints. An optical microscope (OM), Zeiss Axio Scope A1 was used for microstructural examination. Energy dispersive X-ray spectroscopy (EDX) in an environmental scanning electron microscope (SEM), Hitachi S-3400N VPSEM was used to determine chemical compositions of the specimens at an accelerating voltage of 15 kV, an emission current of 76 μA, a working distance of 6.8 mm, an elevation of 35° and a live time of 50 seconds.

40 3.2.5 Welding Quality Acceptance Criteria The quality of welds produced in this experiment were evaluated against a set of welding acceptance criteria from several international standards on welding as illustrated in Table 5. The American Welding Society standard, AWS D17.1:2001 was applied to laser beam welding of both AA 2024-T3 and Ti-6Al-4V, however, the standard was developed for general purpose fusion welding and not specifically for laser beam welding. The European Standard, BS EN ISO 13919-2:2001 was developed specifically for electron beam and laser beam welding of aluminium based alloys but not for titanium based alloys. Another European Standard BS EN 4678:2011 was developed specifically for laser beam welding of metallic materials for aircraft applications and it includes welding quality acceptance criteria for both aluminium and titanium based alloys. It is often the case where for aerospace applications, quality acceptance criteria more stringent than those cited by these standards are required and therefore, the most stringent criteria from each standard were used [227].

Table 5 Relevant welding quality acceptance criteria from AWS D17.1:2001, BS EN ISO 13919-2-2001 and BS EN 4678:2011 standards (t = thickness) [19,231,240]

AWS D17.1:2001 BS EN ISO 13919-2-2001 BS EN 4678:2011 Imperfection Class A Class B Class C D C B Al Ti Cracks Reject Reject Reject Reject Reject Reject Reject Reject ≤ 0.25t, Incomplete Reject Reject Reject max. 1 Reject Reject Reject Reject Fusion mm Incomplete ≤ 0.25t, Reject Reject Reject Reject Reject Reject Reject Penetration max 1 mm ≤ 3 mm + Face Width N/A N/A N/A N/A N/A N/A ≤ t + 1 mm 0.1t ≤ 1 mm + > 1 mm & Root Width N/A N/A N/A N/A N/A N/A 0.5t ≤ 3 + 0.1t 0.33t or 0.50t or ≤ 0.4t, ≤ 0.3t, ≤ 0.5t, >0.2 mm > 0.2 mm Porosity 1.5 mm, 2.3 mm, N/A max. 5 max. 4 max. 6mm & ≤ 0.3t & ≤ 0.2t smaller smaller mm mm 4x larger 2x larger 0.25t, 3x smaller 3x smaller Porosity 0.5t, max. 0.5t, max. adjacent adjacent N/A max. 5 adjacent adjacent Spacing 10 mm 15 mm pore pore mm pore pore 0.015t or 0.025t or ≤ 0.15t, ≤ 0.1t, ≤ 0.05t, Undercut 0.05 mm 0.05 mm, 0.05 mm, max. 2 max. 1.5 max. 1 ≤ 0.15 mm ≤ 0.05t smaller smaller mm mm mm Underfill, 0.015t or 0.025t or ≤ 0.15t, ≤ 0.1t, ≤ 0.05t, 0.1t or 0.5 0.1t or 0.5 Concavity or 0.13 mm 0.13 mm, 0.13 mm, max. 2 max. 1.5 max. 1 mm, mm, Shrinkage greater greater mm mm mm smaller smaller Groove Excess Weld 0.33t or 0.2 mm + 0.2 mm + 0.2 mm + ≤ 0.2t + ≤ 0.15t + Metal or 0.76 mm, N/A N/A 0.3t, max. 0.2t, max. 0.15t, 0.1 mm 0.1 mm Penetration greater 5mm 5mm max. 5mm

Welding imperfections compromise the usefulness of welded joints. Different types of welding imperfections are classified and described in the standard, BS EN ISO 6520, including [241]:

 Cracks: An imperfection produced by a local rupture in the solid state which can arise from the effect of cooling or stresses.  Incomplete fusion and penetration: A lack of union between the weld metal and the parent material.

41  Porosity: A gas cavity formed by entrapped gas of essentially spherical form.  Undercut: An irregular groove at a toe of a run in the parent material or weld metal.  Shrinkage groove: Undercuts visible on each side of the root run.  Underfill: A channel in the surface of a weld due to insufficient deposition of weld metal.  Root concavity: A shallow groove due to shrinkage of a butt weld at the root.  Excess weld metal or penetration: Too large reinforcement of the butt weld on the face or on the root side.

In all three standards, welds containing cracks, incomplete penetration or incomplete fusion are rejected without tolerances. Porosity and undercut especially, can act as stress concentrations so limits are placed in the standards on both weld porosity levels and undercut depth. Undercutting is a common defect in laser beam welding due to its high processing speed which leads to insufficient time for the molten base metal to flow back to the weld toe and fill the gap. Undercut, underfill, root concavity and shrinkage grooves are particularly undesirable for components subject to dynamic loading as these defects act as stress concentrators and subsequently initiate fatigue cracks.

Welding quality acceptance criteria on weld face width and root width are only specified in

AWS D.17. In addition to this, the ratio of root width to face width (Rw) can also be used to evaluate the processing stability of full penetration welding. An investigation conducted by Chen et al. on Nd:YAG laser welding on 3 mm thick 5A90 Al-Li alloy and 2.5 mm thick BT20 titanium alloy proved that a weld width ratio of greater than 0.4 for the titanium alloy and 0.6 for the aluminium alloy produced the conditions for a stable full penetration welding [72]. Therefore, the recommended values of 0.4 and 0.6 were used to evaluate the weld shapes of AA 2024-T3 and Ti-6Al-4V, respectively.

For Ti-6Al-4V, an additional welding quality acceptance criteria on examining the colour of titanium welds is given in BS EN 4678, where a silver glossy metallic grey appearance should be observed in the as-welded state. No white, purplish or blue colours are accepted in the fusion zone (FZ) but a light yellow colour in the FZ or a light blue colour in the heat affected zone (HAZ) is permitted for single pass welding [231]. For all the Ti-6Al-4V welding trials conducted in this investigation, the weld seams remained bright and silvery.

3.3 Results and Discussion on AA 2024-T3 AA 2024-T3 welds were produced under various welding conditions of laser power, welding speed, focal position, filler metal feed rate and shielding gas. All welds were autogenous bead- on-plate welds protected either by industrial grade argon or helium shielding gas with 99.999% purity. The weld quality for each set of welding parameters was assessed using the criteria shown in Table 6.

42 Table 6 Weld quality assessment criteria from Table 5 applied to 3 mm thick AA2024-T3 as defined by AWS D17.1, BS EN ISO 13919-2 and BS EN 4678 [19,231,240]

Face width Root width Porosity Undercut Underfill Reinforcement Standard Level (mm) (mm) (mm) (mm) (mm) (mm) AWS D17.1 Class A N/A N/A ≤ 0.99 ≤ 0.05 ≤ 0.13 ≤ 0.99 BS EN ISO 13919-2 stringent B N/A N/A ≤ 0.90 ≤ 0.15 ≤ 0.15 ≤ 0.65 BS EN 4678 AA ≤ 4.00 ≤ 2.50 ≤ 0.90 ≤ 0.15 ≤ 0.30 ≤ 0.55

The main criteria assessed were the top and bottom weld widths, the ratio of root to face width

(Rw) the depth of undercut and underfill, the size of weld porosity and the height of reinforcement or excess weld penetration. Any weld showing a lack of penetration or crack were rejected according to AWS D17.1, BS EN ISO 13919-2 and BS EN 4678. Other welding imperfections such as overlap and spatter were also identified. The Rw was used to evaluate the processing stability of full penetration welding and a value of 0.6 as determined by Chen et al. [72] from Nd:YAG laser welding on 3 mm thick 5A90 Al-Li was used.

As it can be seen from Figure 6, the operating window was dependent on the welding parameters used. For example, at low laser powers, regardless of the welding speeds chosen, lack of penetration was unavoidable. Lack of penetration is a type of defect which cannot be accepted but it was observed in specimens welded at a laser power of 1.9 kW and a welding speed in the range of 1.0-3.0 m/min. Also, underfill and porosity were often observed in these partially penetrated specimens. However, by increasing the laser power, a wider operating window was obtained [81] and so it was possible to produce a good weld quality by using the right combination of laser power and welding speed. In general, a moderate laser power in the range of 2.5-4.5 kW and a low welding speed in the range of 1.0-2.0 m/min, as shown in Figure 6, produced welds without defects. Increasing the laser power beyond 4.5 kW at a given welding speed led to large longitudinal cracks in the weld due to strong keyhole effects, whereas, decreasing the laser power failed to fully penetrate the parent material. In contrast, increasing the welding speed at a given laser power resulted in the formation of welding defects such as undercut and underfill. Even though Figure 6 shows that a sound weld quality can be obtained with an optimised combination of laser power and welding speed as indicated by the blue circle, there were other parameters such focal position, weld pool shielding and adequate use of filler metal which also had to be considered to avoid welding defects including solidification cracking, grain boundary liquidation and porosity, which were particularly known issues when welding crack sensitive AA 2024-T3 [11]. Therefore, the effect of each parameter was investigated and discussed separately while keeping the rest constant.

43 6

Cracks Undercut 5

Underfill 4

Insufficient Penetration

Laser power Laser power (kW) 2

1 Partial Penetration Full Penetration 0 0 1 2 3 4 5 6 7 8 Welding speed (m/min) Figure 6 Operating welding process parameters window of laser power and welding speed for AA 2024-T3

Figure 7 shows the typical microstructures of a good quality fibre laser beam welded joint, which in this case was produced with a laser power of 4.9 kW, a welding speed of 3.0 m/min, a focal position of +2.0 mm above the top surface of the workpiece, a 1.0 mm diameter 4043 consumable filler metal at a feed rate of 5.1 m/min and argon gas shielding at a flow rate of 20 l/min. The weld exhibited an hourglass shaped transverse cross-section and consisted of fusion zone (FZ), heat affected zone (HAZ) and base metal (BM). The equilibrium precipitate phases of the BM, AA 2024, depended on the weight ratio of Cu and Mg, and were dominantly the semi-coherent CuMgAl2 (S phase) with some CuAl2 (θ phase) [242]. The dark precipitates observed in the BM were the intermetallic compounds, varying in size, shape and chemical composition, formed during the solution heat treatment and natural ageing in T3 temper before welding, some of which included CuMgAl2, CuAl2, Al20Cu2Mn3, Al7Cu2Fe, Al10CuMg and

Al3CuFeMn [26,243]. The HAZ was the region in between the BM and the FZ, which experienced peak temperatures below the melting point of the BM but high enough to exceed the solvus curves, and affect its microstructure.

As it can be seen from Figure 7, the HAZ consisted of two parts, the region near the FZ where the dissolution of the strengthening precipitates occurred and the region next to the BM where over-ageing by coarsening of the strengthening precipitates, semi-coherent Sʹ (CuMgAl2) phase, and then a transformation to the incoherent stable S (CuMgAl2) phase occurred [242,244]. It was reported in other studies that this overaged zone has the maximum amount of the coarser and sparser stable phase S and does not respond well to post-weld ageing [242] without solution heat treatment. The size of the HAZ was very narrow, less than 0.1 mm with only few grains across its width, due to the high thermal gradient, low heat input and fast cooling rate which were characteristics of fibre laser beam welding and therefore, limited metallurgical modifications.

The dissolution of precipitates in the FZ resulted in its softening. The effect was more pronounced in the FZ than the HAZ due to higher temperatures experienced and so a greater dissolution of strengthening phases occurred as a consequence [26,159]. The fast welding speed and also high thermal gradients at the FZ boundary led to an elongated weld pool which promoted formation of columnar dendritic structures. As it can be seen from Figure 7 e), the centre of the FZ showed the formation of characteristic equiaxed dendrites, and from Figure 7 f) columnar dendrites near the FZ boundary. Epitaxial continuous growth of columnar dendrites was observed in the direction of thermal gradients in the FZ, with the same crystallographic orientation to that at the FZ line. The size of the dendrite cells was dependent on the heat input, laser power and welding speed, all of which were responsible for the

45 temperature gradient and the solidification rate. Decreasing the welding speed led to larger dendrite cells in the FZ, wider FZ width and larger grain size in the HAZ [136]. Dendritic growth was reported by Watkins et al. [245] for alloys containing less than 5% weight copper, in this case AA 2024, with the dendrites being α-Al and with either CuAl2 precipitates or CuAl2-Al eutectic as the inter-dendritic phase.

The FZ of AA 2024 mainly consisted of α -Al phase with a surrounding eutectic CuMgAl2 phase as it contained magnesium. The initial solidification occurred epitaxially near the outer FZ boundary with a fine grain zone of planar front growth from the BM. The planar front solidification switched to dendritic grain growth of elongated columnar dendrites towards the centre due to constitutional super-cooling. During welding of AA 2024, the low melting point eutectic with a wide range of freezing temperatures segregated in the grain boundaries and formed the low melting point constituents, which were rejected by the solidifying columnar dendrites. The amount of eutectic liquid between grains were large enough to form a thin, continuous grain boundary film during solidification at a depressed liquidus and solidus temperatures because of Mg, compared to the bulk solidus temperature. The solidus temperature was further suppressed due to a lack of diffusion resulting from rapid non- equilibrium solidification during welding. The shrinkage strains was proportional to the coherence range between the first formation of mushy stage by dendrite interlocking and the solidus, so a wider coherence range increased the tendency for solidification cracking [246]. When the amount of liquid available during the freezing process was insufficient to fill in the spaces between the solidifying grains at the centre, then cracks such as those observed in Figure 10 were formed due to the lack of material and high shrinkage strains in the weld pool [247].

Equiaxed dendritic structure on the other hand, reduced the crack sensitivity due to the abundance of liquid metal between grains which were able to deform more easily under stresses [248], and the lower coherent temperature range resulting from the formation of equiaxed dendrites at a later stage in freezing [34]. In addition, the fine isotropic grain structure of equiaxed dendrites unlike coarse anisotropic columnar dendrites, increased the resistance to crack formation and propagation [12] by distributing the low melting point segregates over a larger grain boundary area and also relieved local shrinkage strains developed during freezing more efficiently [34]. Equiaxed dendrite formation is important for the grain refinement of welds but due to the high solidification rate and thermal gradient, it is often considered difficult to obtain. Instead, columnar dendrite growth is favoured and there is a small chance of equiaxed dendrite formation, resulting in predominantly coarse, low ductility columnar dendritic structure in the FZ [249]. However, since AA 2024 as shown in Table 2, contained a grain refiner titanium which acted as nucleating surfaces, a columnar to equiaxed transition

46 (CET) was promoted and the nucleation of equiaxed dendrites was observed at the weld centreline [250].

The hot cracking sensitivity of AA 2024 in this investigation was therefore, reduced by controlling the weld shape as it dictated the solidification pattern, with a wider weld pool reducing the risk of solidification cracking, shrinkage strain during solidification and the weld metal composition. The weld shape was controlled through optimisation of welding parameters and the use of a filler metal with a different chemistry to that of the base metal to adjust the weld metal chemistry to a low cracking sensitivity range. CET was promoted by the reduction of solidification rate and thermal gradient in the weld pool by using a lower welding speed and a wider weld pool [12] which decreased segregation and stress levels across the weld seam. The use of 4043 filler metal which has a freezing range of around 5°C enabled rapid solidification of welds and reduced the time for shrinkage during solidification.

3.3.1 Effect of Power Density The effect of power density on the weldability and morphology of 3 mm thick AA 2024-T3 specimens was investigated under three different welding conditions of similar nominal heat inputs. The power density at the focal point was determined by the incident power for the 1/e2 beam diameter, the radial distance at which the beam contained 87.5% of the total beam energy. It was larger by 1.32 times for the doughnut shaped TEM01* mode. The TEM01* mode, although it cannot be focused as small a spot diameter as the Gaussian beam, it was used because the power density is more evenly distributed and smaller changes in workpiece position within the focal depth do not affect welding performance, so more stable and consistent welding process was achieved [251]. The power density was varied by changing the laser power at a fixed focused beam size. Both the laser power and the welding speed were varied together to obtain different power densities, with approximately constant heat inputs. All specimens were autogenous bead on plate welds, welded with a positive focal position of +4 mm above the surface of the workpiece and argon gas shielding. The three parameters sets used were: (i) a low laser power of 2.9 kW and a low welding speed of 1.8 m/min, (ii) a moderate laser power of 3.9 kW and a moderate welding speed of 2.4 m/min; and (iii) a high laser power of 4.9 kW and a high welding speed of 3.0 m/min. It was crucial to keep the power density at the workpiece above 106 W/cm2 to achieve and maintain the keyhole mode of welding, by vaporising the material before all the heat dissipates by conduction [115]. It was also necessary to keep the power density below 107 W/cm2 as it could result in the appearance of welding defects such as undercut, underfill, excessive penetration and spatter [252] or even drill of a hole and melt through instead of welding. If less than 105 W/cm2, then the welding mode switches from the keyhole to a conduction limited mode because the laser is only absorbed at the surface and does not penetrate into the workpiece. If even less than

47 103 W/cm2, decoupling of laser energy to workpiece occurs and the material does not melt at all [56].

The power density, as mentioned above, was controlled by changing the laser power. Increasing the power density allowed to reduce the width of the HAZ and weld at a faster welding speed with an increased thermal efficiency. Higher power density can enlarge the operating window [81] but also increase the amount of laser induced plume and loss of alloying elements by vaporisation, encouraging the formation of welding defects [252]. Naeem et al. [253] investigated the influence of power density on porosity formation in welds by performing 4 kW CW Nd:YAG laser welding of AA 5083, 5251 and 6082 sheets. A reduction in porosity was attributed to the increased power density at the workpiece which stabilised the keyhole during the welding process and increased solidification time, allowing the pores to escape [46,253]. Increasing the power density resulted in a greater radiation pressure inside the keyhole, which was responsible for maintaining the keyhole open and as a result, reduced the formation of porosity in the weld. Therefore, the keyhole stability increased with the power density by preventing keyhole collapse and porosity formation. The heat input or line energy associated with the power density used, was determined by dividing the laser power by the welding speed. The laser power and the welding speeds were chosen to produce nominal heat inputs within a very close range between 96.5 and 98.5 J/mm.

Figure 8 shows the face and root widths of the welds produced under three different power densities of 2.05, 2.76 and 3.47 MW/cm2. All specimens passed the BS EN 4678 criteria on the maximum acceptable face and root widths of a laser weld, of 4.0 and 2.5 mm, respectively, as indicated by the blue and red lines. Since full penetration was achieved in all specimens with an adequate top and bottom weld widths, the processing stability of full penetration welding in terms of Rw with a value above 0.6 required for aluminium alloys as indicated by the blue line, were all satisfied. The calculated Rw values were in fact, nearly twice as large, at around 1.2. The maximum height of reinforcement observed in the specimen welded with the moderate power density, of around 0.46 mm was still acceptable when compared to the maximum height of 0.55 mm specified in the most stringent criterion of BS EN 4678.

As it can be seen from Figure 9, the measured depth of undercut was the largest in the specimen welded with the highest power density, at around 0.22 mm. It was greater than the maximum depth of 0.05 mm specified in AWS D17.1 as well as 0.15 mm in BS EN ISO 13919- 2 and BS EN 4678. The high power density of 3.47 MW/cm2, therefore, produced undercut defect which failed the weld quality acceptance criteria. It was determined that the high power density obtained by using a high laser power and welding speed, resulted in excessive levels of undercutting due to an increased amount of parent material being melted and lost through evaporation and spatter, and so the gap during solidification was insufficiently back filled.

48 a) Heat input (J/mm) 96.5 97.0 97.5 98.0 98.5 5 P= 2.9 kW P= 3.9 kW P= 4.9 kW V= 1.8 m/min V= 2.4 m/min V= 3.0 m/min 4

2 Weld width Weldwidth (mm)

1 Top Bottom 0 1.5 2.0 2.5 3.0 3.5 4.0 Power density (x106 W/cm2) b) Heat input (J/mm) 96.5 97 97.5 98 98.5 1.4 0.5 Rw Undercut Reinforcement 1.2 0.4

P= 2.9 kW P= 3.9 kW P= 4.9 kW

1.0 V= 1.8 m/min V= 2.4 m/min V= 3.0 m/min 0.3

w R

0.8 0.2 Imperfections(mm)

0.6 0.1

0.4 0.0 1.5 2.0 2.5 3.0 3.5 4.0 Power density (x106 W/cm2) Figure 8 Influence of power density on a) top and bottom weld widths and b) weld width ratio, undercut and reinforcement

1.0 mm 1.0 mm 1.0 mm 2.05 MW/cm2 2.76 MW/cm2 3.47 MW/cm2 Figure 9 Transverse sections and weld top bead profiles produced with different power densities

49 However, the size of undercuts observed in the other two specimens welded at lower power densities of 2.05 and 2.76 MW/cm2, were both less than 0.15 mm so these specimens passed the criteria in BS EN ISO 13919-2 and BS EN 4678. While the specified undercut depth limit of 0.05 mm as defined in AWS D17.1 was not satisfied by the specimens welded at power densities of 2.05 and 2.76 MW/cm2, such limit was considered very strict and therefore, it was concluded that the measured undercuts were acceptable since they still meet the less stringent criteria. The weld shape of the specimens as illustrated in Figure 9, were hourglass glassed often found in laser welds. ai) aii)

500 μm 150 μm bi) bii)

500 μm 150 μm ci) cii)

500 μm 150 μm

An undercut, as discussed above, was observed both at the top and bottom surfaces of the specimen welded at 3.47 MW/cm2, whereas, only found at the top surface of the specimen welded at 2.05 MW/cm2. Only a small undercut was observed at the top surface of the

50 specimen welded at 2.76 MW/cm2 but a relatively large excessive root penetration was observed.

Despite the acceptable visual appearance and the absence of porosity, intergranular cracks were observed in these welds as shown in Figure 10. Optical microscopic examinations on the FZ of all specimens identified an increasing area of equiaxed dendritic structures with decreasing power density which effectively reduced the solidification rate and the thermal gradients in the weld to promote the growth of equiaxed dendrites. Solidification cracking in the FZ occurred as it solidified due to the influence of tensile residual stresses acting on the low melting point eutectics. The fracture occurred at the grain boundaries and the fractured surface was dendritic with blunt crack tips. Since cracking was observed in all specimens for the given power densities and heat input, it was necessary to investigate alternative methods to reduce solidification cracking.

Low heat input causes increased susceptibility to solidification cracking, whereas, high heat input reduces the cooling rate and increases the time for residual liquid to refill and heal initiated cracks. Although using higher welding speeds and lower heat input result in finer dendrite structure and grain size [36,254], they also cause high thermal shrinkage strains and increase stress gradient that result in a high crack initiate rate [12]. Increasing the heat input further by decreasing the welding speed while maintaining the high power density, would produce wider welds, lower both cooling and solidification rates, and thus minimising thermal tensile strains and reducing crack susceptibility [35]. The crack susceptibility could also be reduced by using a filler metal to control the weld metal composition to a less susceptible level.

The loss of alloying elements such as magnesium with a low boiling point of 1091°C and high equilibrium vapour pressure was measured via energy dispersive spectrometry (EDX) analysis in the transverse cross-section of the weld. The vaporisation of these elements occurs at the surface of the weld pool into the surrounding gas phases [42]. Figure 11 shows the EDX spectrum of the fusion zone obtained from the scanning electron microscopes (SEM) and the analysed chemical composition of the parent material and welds produced under the three power densities. An average loss in magnesium content of around 0.5%, from 1.2 to 0.7% was observed in the welds compared to the parent material (PM), but only small difference was found with respect to the power density. The reduced Mg content in the weld was responsible for keyhole instability which increased the risk of porosity formation and also hot crack susceptibility [42,158,255]. A change in power density was expected to affect the extent of alloying elements losses by influencing the temperature of the molten metal in the weld pool. However, as the nominal heat inputs used among these specimens were similar, the difference was found to be rather small. Nevertheless, the rate of vaporisation and the volume of the weld pool can be controlled by varying welding parameters such as welding speed and

51 laser power. The extent of Mg vaporisation was a function of the welding speed. As the welding speed increased, there was less time to evaporate Mg so high speed reduced elements losses. Vaporisation of the alloying elements was lower when using faster welding speed due to shorter laser and material interaction time so the welding process was more stable. The observed loss of Mg was responsible for keyhole instability and the formation of macro porosity in the weld bead. The effect of welding with filler metal and the associated loss and addition of alloying elements on the weld quality and performance.

100 P=4.9 kW, V=3.0 m/min, f=+4 mm, Ar 90 87.57 86.19 P=3.9 kW, V=2.4 m/min, f=+4 mm, Ar 85.35 83.95 P=2.9 kW, V=1.8 m/min, f=+4 mm, Ar PM 80

Weight(%) 40

10 5.97 5.84 6.36 6.55 0.69 0.63 0.70 1.26 0 Al Cu Mg Figure 11 Weight percentage (%) of chemical elements in base metal and welds of different power densities

3.3.2 Effect of Laser Power The effect of laser power on the weldability and morphology of 3 mm thick AA 2024-T3 specimens was investigated under three different welding conditions. For each condition, only the laser power was changed while keeping all the other parameters constant. All specimens were autogenous bead on plate welds. Various laser powers were employed as listed in Appendix A.

(i) Figure 12 shows the variations in top and bottom weld widths with laser power for the first set. By raising the laser power while keeping the welding speed constant, the weld width increased due to increased heat input and power density which led to more molten metal. A linear relationship between weld width and laser power was found. Both the top and the bottom weld widths at all three laser powers were less than the maximum acceptable widths specified in BS EN 4678 of 4.0 and 2.5 mm, respectively as shown in Table 6. There was no minimum level specified in the standard for the weld widths. The Rw of all specimens were above the minimum value of 0.6 needed to produce a stable full penetration weld. A trend was found where the Rw increased by a small degree with increasing laser power.

52 a) 5

2 Weld width Weldwidth (mm)

1 Top Bot

0 1 2 3 4 5 6 Laser power (kW) b) 1.2 0.50 Rw Underfill Undercut 1.0 Reinforcement 0.40

0.8

0.30 ,

w 0.6 R

0.20 Underfill(mm)

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 1.0 2.0 3.0 4.0 5.0 6.0 Laser power (kW) Figure 12 a) Relationship between weld width and laser power at a welding speed of 2.0 m/min with no defocus and with argon shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.9 kW 2.9 kW 3.9 kW 4.9 kW Figure 13 Transverse sections and weld top bead profiles produced with different laser powers at a welding speed of 2.0 m/min with no defocus and argon shielding gas

53 The size of underfill observed at 1.9 kW, of around 0.42 mm, was significantly greater than almost times the maximum tolerated depth of 0.13 mm specified in AWS D17.1 and 0.15 mm in BS EN ISO 13919-2 and BS EN 4678. Since a constant beam diameter was used and no underfill defect was observed at higher laser powers, it was suspected that the laser power of 1.9 kW was insufficient to meet the minimum threshold power for full penetration.

A complete penetration of the 3 mm base metal was obtained at laser powers equal to or greater than 2.9 kW. At higher laser powers or heat input, evaporation and expulsion of the weld metal resulted in undercut defects at the toe of the parent material and reinforcement at the weld centre. The height of reinforcement at above 2.9 kW was less than the limits of 0.55, 0.65 and 0.99 mm specified in BS EN 4678, BS EN ISO 13919-2 and AWS D17.1, respectively. The effect of laser power on excess weld metal was therefore, negligible for this set. Even though reinforcement has only small influence on weld quality and performance, it is still important to limit its height because it represents non-value added cost. Undercut defect was more serious at 3.9 kW, where the depth of the undercut was greater than 0.15 mm and thus, failed all criteria. The undercut was smaller in specimens welded at 2.9 and 4.9 kW, less than 0.15 mm but still greater than 0.05 mm as specified in AWS D17.1. Depending on the applications, the undercut defects present in the specimens welded at 2.9 and 4.9 kW may be acceptable but not 3.9 kW. Undercut usually appears due to low viscosity of molten weld pool and can be resolved by using filler wire and improving face and root shielding.

As Figure 13 shows, the underfill defect in the specimen welded at 1.9 kW was clearly visible by visual inspection. The weld quality was far from acceptable. At laser powers greater or equal to 2.9 kW, the weld shape was hourglass shaped. The weld quality of the specimen welded at 2.9 kW was good because only small reinforcement and undercut were formed and there was no underfill defect. Root concavity was observed at 3.9 kW but acceptable according to BS EN 4678 and BS EN ISO 13919-2. The specimen welded at 4.9 kW seemed visually acceptable but was slightly asymmetrical with an unacceptable size of undercut on one side of the weld toe. Therefore, for this set of welding parameters, it was determined that the laser powers of 2.9 and 4.9 kW provided the optimum welding conditions.

(ii) Figure 14 shows the effect of laser power at a faster welding speed 3.0 m/min, while maintaining the same focal position and shielding. It was found that laser power is an important welding parameter for full penetration and to control the weld shape. Sufficiently high power density was required to achieve keyhole welding and control the formation of welds.

54 a) 5

2 Weld width Weldwidth (mm)

1 Top Bot

0 1 2 3 4 5 6 Laser power (kW) b)

1.2 0.50 Rw Undercut Reinforcement 1.0 Underfill 0.40 0.17 mm 0.8 porosity 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 1.0 2.0 3.0 4.0 5.0 6.0 Laser power (kW) Figure 14 a) Relationship between weld width and laser power at a welding speed of 3.0 m/min with no defocus and with argon shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.9 kW 2.9 kW 3.9 kW 4.9 kW Figure 15 Transverse sections and weld top bead profiles produced with different laser powers at a welding speed of 3.0 m/min with no defocus and argon shielding gas

55 A laser power of 1.9 kW was too low for this welding speed, so a lack of penetration was observed, which according to all standards in Table 2, was not acceptable. The significant lack of fusion observed in the previous specimen welded with same laser power of 1.9 kW but a lower welding speed of 2.0 m/min was not solved by increased the welding speed, but instead created another problem related to a lack of penetration, caused by insufficient beam power and heat input. If any aluminium oxide was present on the workpiece, which has a much higher melting point than aluminium, then it could also be responsible for the incomplete fusion [26,157]. However, cleaning of both sides of the joint was done prior to welding so the contribution from contaminations should considered to be relatively small.

The top width of the specimen welded at 1.9 kW was acceptable according to BS EN 4678 but not the bottom width due to lack of penetration. The top and the bottom widths of the specimens welded at higher laser powers were very similar, as indicated by the Rw value being close to 1. The specimens welded at 3.9 and 4.9 kW showed Rw values greater than 1, which meant that the bottom width became slightly larger than the top width. The top width was acceptable for all laser powers but the bottom width at 4.9 kW was very close to the maximum root width limit of 2.50 mm specified in BS EN 4678. This trend can be clearly observed in Figure 14, where the weld shape at 1.9 kW is V shaped, which at 2.9 kW is nearly rectangular and hourglass shaped at 3.9 and 4.9 kW with a wider bottom. Both undercut and underfill defects were found in the specimen welded at 1.9 kW. The undercut was greater than the maximum size of 0.05 mm in AWS D17.1 and 0.15mm in the other two BS standards. The depth of underfill when compared to the specimen welded at 2.0 m/min, was much smaller and passed all criteria. A small amount of undercut, reinforcement and underfill were observed in the specimen welded at 3.9 kW. The underfill and reinforcement defects passed all criteria, whereas, the undercut defect passed the criteria in BS EN ISO 13919-2 and BS EN 4678 but was slightly larger than the 0.05 mm specified in AWS D17.1. No undercut and underfill defects were observed in the specimens welded at 2.9 and 4.9 kW but larger reinforcements were produced compared to the remaining specimens but still satisfied the limits of 0.55, 0.65 and 0.99 mm.

Low heat input at 1.9 kW resulted in porosity at the lower half of the weld with a pore diameter of 0.22 mm as well as a surface porosity as observed by the weld top bead profile in Figure 15. The size of the measured pore was still much smaller than the most stringent requirement of 0.90 mm in BS EN ISO 1391-2 and BS EN 4678 as well as the less stringent 0.99 mm in AWS D17.1. As discussed by Alfieri et al. [26,256], increasing amount of porosity was observed in 3.2 mm thick AA 2024 welded using a disc laser at around 15% porosity content with a lower thermal input of 80 J/mm compared to around 5% porosity content in the FZ with 200 J/mm. In this case, the net heat input at 1.9 kW was only 38 J/mm, whereas, at 4.9 kW

56 was 98 J/mm at the welding speed of 3.0 m/min. If increasing the heat input does not produce full penetration welds then it would only result in enhanced vaporisation of alloying elements which favours keyhole instability and gas occlusions [26,256]. The likely source of micro porosity with a diameter less than 0.20 mm [257], is hydrogen dissolved in the weld pool considering its small spherical shape and would not result in rejection of the welded joint. Entrapment of gas bubbles occur when the solidification rate is too fast and the rejected hydrogen at the solid-liquid interface, which exceeds the low solid solubility limit, does not escape the solidifying weld. The solubility of hydrogen is high in the liquid phase about 20 times greater than in solid but drops significantly during cooling in the solid state [255,258], from around 0.65 ml/100g in molten pure aluminium to 0.034 ml/100g in solid aluminium [259]. Another mechanism for the formation of spherical pores is the inclusion of shielding gas, which is more frequently observed in pulsed wave laser beam welding [260]. The key method of reducing porosity was to remove hydrogen sources before and during welding or to produce a hydrogen oversaturated weld by increasing the solidification rate, for example, by increasing the welding speed. Except for the specimen welded at 1.9 kW, the weld quality of the specimens welded at greater or equal to 2.9kW were all acceptable.

(iii) Identical welding parameters to those of the first set were used for the third set but with a +4 mm positive defocusing. Power density is the maximum at the focal point. It can be reduced by focusing the beam below or above the workpiece, leading to a wider weld pool and larger keyhole, often considered to be beneficial for process stability due to reduced risk of solidification cracking. Figure 17 shows that the specimen welded at 1.9 kW has a zero root width (incomplete penetration) just like those from the previous sets with the same laser power of 1.9 kW. The top weld width of this specimen however, was approximately 0.5 mm wider than without defocusing. The weld shape of the specimens welded at higher laser powers, changed due to widening of the top weld width caused by a widened intensity profile of the laser beam. The weld shape at 1.9 kW was V shaped with an Rw of 0, nail head shaped at 2.9 kW with an Rw of around 0.8, and hourglass shaped at 3.9 kW and 4.9 kW with Rw of around

0.9 and 1.1, respectively. The Rw value greater than 1 at 4.9 kW indicated that the bottom width was larger than the top width as shown in Figure 16.

Excessive weld metal or reinforcement was observed in all specimens but all less than the maximum height of reinforcement allowed in all three standards. Undercut defect was only observed in the specimen welded at 3.9 kW which passed both the BS EN ISO 13919-2 and BS EN 4678 criteria of 0.15 mm but failed the AWS D17.1 criterion of 0.05 mm.

57 a) 5

2 Weld width Weldwidth (mm)

1 Top Bot

0 1 2 3 4 5 6 Laser power (kW) b) 1.2 0.50 Rw Undercut Underfill 1.0 Reinforcement 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 1.0 2.0 3.0 4.0 5.0 6.0 Laser power (kW) Figure 16 a) Relationship between weld width and laser power at a welding speed of 2.0 m/min with +4 mm defocus and with argon shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement

1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.9 kW 2.9 kW 3.9 kW 4.9 kW Figure 17 Transverse sections and weld top bead profiles produced with different laser powers at a welding speed of 2.0 m/min with +4 mm defocus and argon shielding gas

58 A large underfill was identified in the specimen welded at 2.9 kW of almost 0.5 mm, which was considerably greater than the least stringent maximum limit of 0.30 mm in BS EN 4678 as well as the more stringent 0.13 mm in AWS D17.1 and 0.15 mm in BS EN ISO 13919-2. A smaller underfill in the specimen welded at 4.9 kW compared to that at 3.9 kW was detected, which failed the AWS D17.1 and BS EN ISO 13919-2 criteria but passed the BS EN 4678 criterion. However, the impact of undercut and underfill defects on the weld quality of the specimens welded at 3.9 and 4.9 kW was small compared to the longitudinal cracks located at the weld centre. Increasing the focal position to +4 mm resulted in the root widths increasing to a level which was very close to the maximum limit specified in BS EN 4678 of 2.50 mm. Therefore, rather than improving the keyhole stability, the opposite happened where central solidification cracking occurred in the specimens welded at laser powers of 3.9 and 4.9 kW, because the weld pool formed during welding became too wide. Consequently, the availability of liquid metal in the weld pool was not enough to back-fill the empty regions opened by shrinkage strains between solidifying weld metal and so a crack was formed. Since increasing the laser power from 3.9 to 4.9 kW led to more material being melted, the contraction strain became larger and caused larger cracks to form. Also, low solidification rate due to high laser power cause the impurity elements to diffuse into the molten pool and the resulting weakened microstructure of the weld in the middle was susceptible for cracking [261]. As it can be seen from the cross-sections in Figure 17, cracks are located in the middle of the weld seam.

Porosity was observed in the two specimens welded with lower laser powers of 1.9 and 2.9 kW. The pore diameter was only 0.023 mm at 1.9 kW and 0.092 mm at 2.9 kW, both of which were at least an order of magnitude smaller than the maximum accepted pore diameter of more stringent 0.90 and less stringent 0.99 mm specified in the standards. Such pores were attributed to hydrogen porosity owing to their small size. It was therefore concluded, under the welding conditions used in the third set with a welding speed of 2.0 m/min, +4 mm defocus and argon shielding, the overall weld quality was unacceptable for all laser powers investigated, from 1.9 to 4.9 kW.

3.3.3 Effect of Welding Speed The effect of welding speed was investigated under three different welding conditions. The welding speed was changed while the other parameters were kept constant Various welding speeds were employed as listed in Appendix A.

Welding speed was found to influence the weld shape and weldability [220–222], The weld width was inversely proportional to the welding speed for a given power so a low welding speed led to an increased heat input and a wider weld width. High welding speed was found to reduce the processing stability and induced welding defects such as porosity, undercut and underfill.

59 (i) Figure 18 shows the relationship between top width and welding speed for the specimens welded without filler metal at a laser power of 1.9 kW, +4 mm positive defocusing above the surface of the workpiece and with argon shielding gas. It was found that an increase in welding speed led to a decrease in the top weld width due to reduced heat input, meaning less material being melted.

It was previously discussed in Section 3.3.2 that for the same welding parameters, the weld quality was poor at 1.9 kW, so the possibility of welding at 1.9 kW was further investigated here by welding at lower welding speeds in the range between 1.0 and 2.0 m/min. By decreasing the welding speed, the heat input increased, making the weld width wider. However, full penetration was still not achieved meaning that it was not possible to suppress lack of penetration even by decreasing the welding speed down to 1.0 m/min at 1.9 kW. It was most likely that the minimum threshold power density was not achieved to perform full penetration welding. The determined power density at 1.9 kW was 0.672 MW/cm2 at focus which was less than 1 MW/cm2 required to form a keyhole. As it can be seen from the weld shape of the specimens in Figure 19, the depth to width ratio is close to one, similar to those commonly obtained in conduction mode.

No undercut defect was produced but underfill was observed, especially at 1.0 and 1.7 m/min. The depth of underfill at 1.7 m/min was greater than the criteria of AWS D17.1 and BSEN ISO 13919-2 as shown in Table 6, which are 0.13 and 0.15, respectively, but less than the criterion of BS EN 4678 which is 0.30 mm. Due to the fact that full penetration was not achieved at all welding speeds, the Rw was zero for all specimens. A large macro pore as shown in Figure 3 was observed in the specimen welded at 1.0 m/min. The diameter of the pore was 0.592 mm, whose size was in the range between 0.3 and 0.6 mm so it was classified as a macro pore. Macro pores unlike micro pores are formed due to keyhole collapse during welding because of large differences in melting and boiling points of the parent material and the alloying elements such as magnesium in AA 2024 [26]. The vaporisation of magnesium during welding increases the tendency to form macro pores by affecting the keyhole pressure [262], which in this case was controlled by the welding speed. Partial penetration welds are more prone to porosity than fully penetration welds because the path for the escape of gas bubbles is only available via the top surface, whereas, in full penetration welds, the pores which form at the root of the weld can escape from the bottom surface instead of travelling up to the top surface.

60 a) 5

2 Weld width Weldwidth (mm)

Top

0 0.5 1.0 1.5 2.0 2.5 Speed (m/min) b) 1.2 0.50 Underfill Reinforcement 1.0 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 0.5 1.0 1.5 2.0 2.5 Speed (m/min) Figure 18 a) Relationship between weld width and welding speed at a laser power of 1.9 kW with +4 mm defocus and with argon shielding gas, and b) the resultant underfill and reinforcement defects

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 m/min 1.3 m/min 1.5 m/min 1.7 m/min 2.0 m/min Figure 19 Transverse sections and weld top bead profiles produced with different welding speeds at a laser power of 1.9 kW with +4 mm defocus and argon shielding gas

61 Porosity was not observed at higher welding speeds, either because the keyhole stability increased with welding speed or there was not enough time for nucleation of porosity. The largest pore size of 0.592 mm measured in the specimen welded at 1.0 m/min, still passed the maximum size of 0.90 mm in BS EN ISO 13919-2 and BS EN 4678, and 0.99 mm in AWS D17.1. However, in accordance with these standards, the weld quality achieved using 1.9 kW laser power at a speed of 1.0-2.0 m/min and +4 mm defocus was rejected mainly due to lack of penetration.

(ii) As it was not possible to achieve full penetration in the welds at 1.9 kW, a further investigation was conducted at a higher laser power of 2.9 kW while maintaining the +4 mm defocus and argon shielding gas. Figure 20 shows that it was possible to achieve full penetration by increasing the laser power to 2.9 kW while using the same welding speeds as those at 1.9 kW. It was found that the weld width increased with decreasing welding speed. The variation of the welding speed affected the heat input and thereby the whole geometry of the joint. The weld width increased significantly by lowering the speed. This widening of the weld width for lower welding speed suggested more heat conduction perpendicular to the welded joint with decreasing welding speed.

The face and root widths of the specimens welded at up to 1.5 m/min were too wide according to the criteria in BS EN 4678 of 4.0 and 2.5 mm for the top and the bottom width, respectively. Such wide weld widths, larger than the thickness of the welded sheet, was not desired as it can have detrimental effects on the weld strength. The specimens welded at 1.7 and 2.0 m/min on the other hand, passed the criteria in BS EN 4678 because their top and bottom weld widths were below the maximum limits. The Rw of all specimens except at 1.7 m/min was greater than 0.6, of around 0.7 so at these welding speeds, the full penetration welding process was stable. The Rw at 1.7 m/min was below 0.6, of around 0.5 which indicated that the processing stability was low for full penetration welding [188] because the root width compared to the face width was much narrower.

Reinforcement in these specimens was mainly identified as excessive root penetration rather than excess weld metal on the top surface. The height of reinforcement was less than the most stringent 0.55 mm cited in BS EN 4678, for all specimens. The fastest welding speed of 2.0 m/min produced an underfill with a depth of around 0.48 mm due to insufficient weld metal. This depth was greater than the maximum depth of 0.13 mm specified in AWS D17.1, 0.15 mm in BS EN ISO 13919-2 and 0.30 mm in BS EN 4678, meaning that it failed all criteria. The fast cooling rate at this welding speed resulted in not enough time for the weld to solidify with a sufficient liquid metal flow and formed an underfill defect [255]. Underfill can act as a crack initiation point and reduce the cross-sectional thickness, which has a negative impact on the weld mechanical properties.

62 a) 5

2 Weld width Weldwidth (mm)

1 Top Bot 0 0.5 1.0 1.5 2.0 2.5 Speed (m/min) b) 1.2 0.50 Rw Undercut Underfill 1.0 Reinforcement 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 0.5 1.0 1.5 2.0 2.5 Speed (m/min) Figure 20 a) Relationship between weld width and welding speed at a laser power of 2.9 kW with +4 mm defocus and with argon shielding gas, and b) the resultant Rw, undercut, underfill and reinforcement defects

1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.1 m/min 1.3 m/min 1.5 m/min 1.7 m/min 2.0 m/min Figure 21 Transverse sections and weld top bead profiles produced with different welding speeds at a laser power of 2.9 kW with +4 mm defocus and argon shielding gas

63 It was not observed in the specimens welded at slower welding speeds because the solidification rate was lower and so the weld seam was sufficiently filled [46]. The depth of undercut measured in the specimens welded at 1.5 and 1.7 m/min failed the maximum depth of 0.05 mm specified in AWS D17.1 but just passed the criteria in the two standards, of 0.15 mm. The size of undercut observed in the specimen welded at 1.3 m/min passed all criteria.

The weld shape was influenced by the welding speed as shown in Figure 21. A lower welding speed resulted in a considerable increase in the weld width due to the increased heat input. The difference between top and bottom weld widths became wider as the welding speed increased meaning that the weld shape change from a rectangular shape with flat edges to a deep narrow V shape with a single weld centreline boundary. The top weld bead profile was relatively flat in all specimens up to 1.7 m/min and then a large underfill at 2.0 m/min. A flat top surface geometry observed in the specimens was an indication of a good weld quality because it meant that there was no weak point acting as a source of stress concentration in the weld seam. Although comparatively wider weld widths close to the maximum limits specified in BS EN 4678 were produced at a welding speed of 1.5 m/min compared to those at 1.7 and 2.0 m/min, the weld quality was good with no underfill and reasonably small undercut and RW above 0.6 so it was concluded that the best weld quality was obtained at the welding speed of 1.5 m/min given the laser power of 2.9 kW and + 4mm defocus.

The weld centre of the specimens in Figure 21 consisted of equiaxed dendrites and fine columnar dendrites close to the fusion boundary. The amount of equiaxed dendrites decreased with increasing welding speed. Fast welding speed promoted a fine dendritic structure and grain sizes resulting from low heat input. The columnar grains growth was epitaxial at higher welding speeds. In contrast, at lower welding speeds, the columnar grains curved away from the normal to the welding direction and instead, aligned parallel to the welding direction.

(iii) At a lower laser power of 2.9 kW, the welding speed available for full penetration welding was limited only to a low speed range, which in general, is not acceptable from the standpoint of productivity. Therefore, the effect of welding speed was investigated at even higher laser power of 4.9 kW at welding speeds greater than 2.0 m/min up to 6.0 m/min, as it was found in Section 3.3.2 that a welding speed of 2.0 m/min was too low at +4 mm defocus, the process became unstable and produced cracks in the weld due to excessing heat input into the molten weld pool [46]. Figure 22 shows that full penetration was obtained using laser power of 4.9 kW at all welding speeds. It was found that at an increased welding speed of 3.0 m/min and above, weld centreline cracking, which was previously observed in the specimen welded at 2.0 m/min, was not produced. Increasing the welding speed decreased the cracking susceptibility through

64 grain refinement and the formation of a finer dendrite structure [263]. The dendrite arm spacing decreased with increasing welding speed so the weld metal microstructure became finer at higher welding speeds and faster cooling rate.

On the other hand, the occurrence of hot cracking is also possible at high welding speeds because the rapid solidification rate at high welding speeds can quickly induce thermal shrinkage strains and increase stress gradient which results in high crack initiation rate, and also reduce time for residual liquid along grain boundaries to back-fill the initiated cracks [12]. It is necessary to use a low welding speed together with a low heat input to minimise the risk of transverse cracking as a result of elongated temperature distribution in the welding direction [264]. Wider welds were generally less crack sensitive than narrow welds so a lower welding speed was recommended.

It was found that the weld widths decreased linearly with increasing welding speed as shown in Figure 22. Increasing the welding speed helped to maintain a stable keyhole welding. The face and the root widths were acceptable in the specimens welded at welding speeds greater than 2.0 m/min. The top width for all specimens was less than the limit of 4.0 mm in BS EN 4678 and also the bottom width was less than the limit of 2.5 mm, except at 2.0 m/min, so by increasing the welding speed at the given laser power, acceptable weld widths were obtained.

The Rw was above 0.6 at all welding speeds meaning that the process stability for full penetration welding was met. Reinforcement found in all specimens was below the limit of 0.55 mm in BS EN 467, 0.65 mm in BS EN ISO 13919-2 and 0.99 mm in AWS D17.1 and no clear correlation between the height of reinforcement and welding speed was identified. Underfill was only observed in the specimen welded at 2.0 m/min, whereas, undercut was observed at higher welding speeds of 3.0 m/min and 6.0 m/min.

An undercut was formed as a groove in the base metal along the edges of the weld as shown in Figure 23 at higher welding speeds. This was because during welding the molten liquid metal from the base metal was drawn along the edges of the weld bead by surface tension and piled up along the weld centre and prevented from wetting back during the rapid solidification. The depth of undercut formed in these two specimens were fairly large with a size greater than the less stringent criteria of 0.15 mm in BS EN ISO 13919-2 and BS EN 4678, as well as the depth of 0.05 mm in AWS D17.1.

65 a) 5 Top Bot 4

2 Weld width Weldwidth (mm)

0 1 2 3 4 5 6 7 Speed (m/min) b) 1.2 0.50 Rw Undercut Underfill 1.0 Reinforcement 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 1 2 3 4 5 6 7 Speed (m/min) Figure 22 a) Relationship between weld width and welding speed at a laser power of 4.9 kW with +4 mm defocus and with argon shielding gas, and b) the resultant Rw, undercut, underfill and reinforcement defects

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm 2.0 m/min 3.0 m/min 4.0 m/min 5.0 m/min 6.0 m/min Figure 23 Transverse sections and weld top bead profiles produced with different welding speeds at a laser power of 4.9 kW with +4 mm defocus and argon shielding gas

66 On the other hand, neither undercut nor underfill defects were detected in the specimens welded at 4.0 and 5.0 m/min, where the top weld geometry was flatter. However, a small micro pore related to hydrogen in the weld pool, was located close to the weld centreline with a diameter of 0.064 mm at 5.0 m/min. It was determined that its influence on the weld quality and performance was small according to the three standards which specify a maximum of 0.90 and 0.99 mm, for the rejection of porosity defects. By increasing the welding speed further to 6.0 m/min, hydrogen porosity was effectively removed because there was not enough time for nucleation of hydrogen pores due to rapid cooling and solidification rate at higher welding speed. As the welding speed increased, the critical hydrogen concentration required to form was greater [255]. It was therefore, determined that the critical welding speed at which the formation and growth of hydrogen porosity is prevented, was above 5.0 m/min in these specimens.

As the welding speed increased, the weld width decreased because the heat input was small due to less exposure of the laser beam on the workpiece at high welding speeds, so the depth to width ratio of the specimens welded at 4.0 and 5.0 m/min was large and the FZ size was minimised. The weld shape at low welding speeds was an hourglass shaped but as the welding speed increased, the shape gradually became straighter with flat edges especially above 4.0 m/min. In summary, at a high laser power of 4.9 kW and with a +4 mm positive defocusing, an improved weld quality was achieved by increasing the welding speed. In particular, at 4.0 m/min, the weld quality was optimised with the minimum number of welding imperfections.

3.3.4 Effect of Focal Position The effect of focal position on the weldability and morphology of 3 mm thick AA 2024-T3 specimens was investigated by changing the defocusing distance. The focal position was changed while all the other parameters were kept constant Various focal positions were employed as listed in Appendix A.

The focal position is defined as the distance of the focal point from the top surface of the workpiece. The location of a focal point determines the incident beam size or the power density, and this affects the weld width and depth of penetration. Therefore, it has to be carefully selected to provide the required power density at which keyhole model welding is realised. The keyhole is formed when the material at the interaction point melts and vaporises. The vapour recoil pressure creates and sustains the keyhole, forming a thin cylinder of molten metal which solidifies to produce a narrow FZ. This means that the focal position influences not only the absorption of the laser beam on the workpiece surface, but also the incidence direction of the beam, so it affects the penetration depth and the weld shape [25]. Variations

67 in the specific energy provided to the workpiece and the size of the FZ as functions of the focal position, results in different material responses [26].

The maximum power density is achieved with a focused beam at 0 mm, where the heat is localised into a small affected area and control the volume of molten material [213]. The power density decreases along either of the beam axis directions away from the focal point. Defocusing leads to a wider melt pool and a larger keyhole which can sometimes benefit the processing stability [82]. In some cases, the laser welding needs some defocusing, because too high power density of the beam at focus can lead to excessive vaporisation of material. However, defocusing the beam reduces the power density such that the amount of laser power that is reflected from the surface is increased and results in fluctuations in the energy absorbed by the workpiece. Using low wavelength laser sources such Nd:YAG and fibres lasers helps to minimise this fluctuation since their wavelength is more efficiently absorbed than the wavelength of CO2 laser.

Negative defocusing allows welding speed to increase at the expense of weld quality, because it improves laser beam to material coupling and irradiance inside the weld pool [142]. Negative defocusing reduces porosity content and enhances penetration depth compared to positive defocusing in disc laser welded AA 2024 [158]. This is because the reflected beam loss is reduced by multiple reflections at the keyhole wall. Refocusing of these reflected beams from the keyhole wall increases the power density which generates a greater vapour pressure at the bottom of the keyhole and as a result, increases beam absorptivity but may also result in temporal fluctuations and spatial instability of the keyhole [25]. Positive defocusing on the other hand, allows positioning of the focal point above the top surface of the workpiece and even on the surface of the filler metal [82]. Plasma formation is increased by positive defocusing, which defocused the beam and reduced the irradiance on the surface but the keyhole was inherently more stable with less porosity formation [255]. The effect of reduction in specific energy due to positive defocusing produced a lower average temperature in the weld pool, and so the vaporisation of low boiling point alloying elements such as magnesium was reduced. In contrast, when fibre laser welding aluminium alloys, the effect of positive defocusing on plasma shielding, effect dominated by inverse bremsstrahlung absorption, is negligible at low laser powers and only appears above 5 kW [172].

Due to the high reflectivity of AA 2024-T3, welding with a focused beam located at the top surface can lead to overheating of the melt pool and reduce the weld quality [82]. Published literature suggest that defocusing the laser beam can improve the weld quality in aluminium joints in term of porosity content and enhanced penetration depth [158,255,257,258]. For example, Alfieri et al. [26] studied the effect of different focal positions on the quality of butt welds produced from 3.2 mm AA 2024 sheet using a disc laser at a laser power of 1.8 kW and

68 a welding speed of 10 mm/s. A negative defocusing of -1.0 mm defocus enabled keyhole mode but at a positive defocusing of +1.0 mm changed to conduction mode. At a lower focal position of +0.5 mm, several porosities were produced due to a lack of penetration [26,158].

The threshold power density for keyhole formation was found to be lower at negative defocusing than at positive defocusing, so the weld pool size at negative defocusing was larger than that for the same magnitude of positive defocusing [255]. The optimum defocusing condition for suppressing porosity formation was determined by considering the wall-focusing effect mentioned above and the geometrical features of the weld pool. Excessive vapour pressure generated by beam concentration effect at the bottom of the keyhole was relieved by adequate control of the defocus value. The focal position was found to be dependent on the amount of alloying elements such as magnesium and zinc, having a low boiling point as well as high vapour pressure. Since AA 2024-T3 contains magnesium and zinc as shown in Table 2, it was expected that focal position has an influence on the weld quality.

(i) According to the results from Section 3.3.2, it was found that for the specimen welded at a laser power of 3.9 kW, a welding speed of 2 m/min and at a focal position of 0 mm, the size of the undercut formed was too big. Therefore, the effect of changing focal position was investigated for the above laser power and welding speed to see if the weld quality could be improved by defocusing. Figure 24 shows that the top and bottom weld widths increased with both positive and negative defocusing. The reason for a wider root width at negative focal positions than at equivalent positive focal positions was attributed to the lower threshold power density for keyhole formation at negative focal positions, due to more efficient laser and material interaction [255].

The top width at all focal positions passed the criterion on the maximum face width of 4.0 mm specified in BS EN 4678, but the bottom width only passed at the focal positions of 0 and +2 mm, below 2 mm wide.

The Rw was above 0.6 at all focal positions. Figure 25 shows that full penetration was achieved in all specimens. It was found that the Rw at negative focal positions was higher than at zero or positive positions with values close to 1, indicating that the top and the bottom weld widths were similar probably due to the beam concentration effect at the lower half of the weld discussed above. The undercut defect obtained at 0 mm was successfully reduced by defocusing the beam and the optimum weld was produced at 2 mm where no undercut was formed.

69 a) 5 P = 3.9 kW, V = 2 m/min, Ar

2 Weld width (mm)width Weld

1 Top Bottom 0 -6 -4 -2 0 2 4 6 Focal position (mm) b) 1.2 0.50 Rw Undercut Underfill 1.0 Reinforcement 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 Focal position (mm) Figure 24 a) Relationship between weld width and focal position at a laser power of 3.9 kW, a welding speed of 2.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm -4 mm -2mm +0 mm +2 mm +4 mm Figure 25 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 3.9 kW, a welding speed of 2.0 m/min and with argon shielding gas

70 On the other hand, the size of undercut at -2 mm was similar to that at 0 mm, greater than the depth of 0.15 mm specified in the less stringent BS EN 13919-2 and BS EN 4678 criteria. The undercut at -4 and +4 mm passed the above criteria but not the depth of 0.05 mm in AWS D17.1. Reinforcement was observed in all specimens except at -2 mm where an underfill was formed instead, but below the maximum limit of 0.55, 0.65 and 0.99 as listed in Table 6. The underfill defect at -2 mm was larger than the maximum limit of 0.13, 0.15 and 0.30 mm.

The weld shape at focus was relatively flat with straight fusion boundaries. As the focal position increased in magnitude, the weld shape increasingly became more hourglass shaped with curved fusion boundaries. It was believed that the weld centreline cracking observed at the focal position of +4 mm was due to the large face width of the weld formed by an increased amount of parent material being melted and lost through evaporation. As a consequence, the gap formed during welding was insufficiently filled back to heal the initiated crack. Also, the reduced power density was thought to increase susceptibility to solidification cracking. Therefore, by taking into account the results from Figure 24 and Figure 25, the optimum weld quality was produced at the focal position of +2 mm.

(ii) The effect of focal position on the weld quality was investigated also with another set of welding parameters using a laser power of 4.9 kW and welding speed of 3.0 m/min. It was previously found in Section 3.3.2 that a good weld quality can be obtained without defocusing, whereas, at a focal position of +4 mm, a large undercut was formed, so the aim was to investigate the behaviour at other focal positions.

Figure 26 shows that the change in the bottom width with focal position was negligible, whereas, the top width increased with increasing focal position. Both the top and the bottom width at all focal positions for the given laser power and welding speed were below the maximum limit of 4.0 and 2.5 mm specified in BS EN 4678, respectively.

The Rw was very consistent for all focal positions of around 0.9 so the processing stability for full penetration was high for all specimens. The weld quality improved with decreasing focal position from +4 mm to +2 mm in terms of undercut defect which failed all criteria at +4 mm but passed the criteria of 0.15 mm in BS EN ISO 13919-2 and BS EN 4678 at +2 mm. Negative defocusing of the beam led to further improvements in the weld quality at -2 mm and -4 mm focal positions where underfill defect was absent and also undercut passed all criteria, even below the most stringent depth of 0.05 mm in AWS D17.1. Reinforcement was present in all specimens but was acceptable at all focal positions according to all criteria. Underfill defect observed at 0 mm, as discussed previously was close to the limit of the most stringent depth of 0.13 mm.

71 a) 5 P = 4.9 kW, V = 3 m/min, Ar

2 Weld width (mm)width Weld

1 Top Bottom 0 -6 -4 -2 0 2 4 6 Focal position (mm) b) 1.2 0.50 Rw Undercut Underfill 1.0 Reinforcement 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections(mm)

0.10 0.2

0.0 0.00 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 Focal position (mm) Figure 26 a) Relationship between weld width and focal position at a laser power of 4.9 kW, a welding speed of 3.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm

-4 mm -2mm +0 mm +2 mm +4 mm Figure 27 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 4.9 kW, a welding speed of 3.0 m/min and with argon shielding gas

72 As shown in Figure 27, no cracking or porosity was found at all focal positions. The weld shape was hourglass shaped at all focal positions with curved fusion boundaries and relatively flat top weld surfaces. Some excessive penetration was observed at 0 mm but still with a good enough weld quality in accordance with the three standards. The most acceptable weld profile was obtained at -2 mm, but also very good at other focal positions except at +4 mm, because of large undercuts and root shrinkage grooves. It was therefore, possible to weld AA 2024-T3 at a high laser power and welding speed, over a wide range of focal positions, with characteristics and qualities that are pertinent to aerospace applications.

(iii) The effect of focal position was investigated for the same combination of welding parameters but with the addition of a 1.0 mm diameter 4043 filler wire into the leading edge of the weld pool at a feed rate of 5.0 m/min. The interaction between the filler metal and the laser beam was controlled by the focal position. When the laser beam was focused to the top surface of the workpiece, 0 mm, or negatively defocused below the surface, then the interaction occurred outside of the focal point where the power density and the absorption of the beam energy were lower, so the filler metal melted not by a direct contact with the laser beam but instead via heating inside the weld pool [265]. In addition, due to the fast welding speed and filler metal feed rate, there was not enough time for the beam to heat the filler wire during wire feed in open air before entering the weld pool. As a result, the filler metal reflected some part of the laser beam instead because the focal position was not optimised with respect to the wire feed when positioned below or at the surface of the workpiece. The fraction of the beam energy reflected from filler metal was minimised by adjusting the focal position in such a way that the filler wire was located at the laser beam impingement point. Positive defocusing allowed the laser beam to impinge the filler wire, minimising the reflected part of the laser power and increasing the amount absorbed [266].

Unlike the trend observed in Figure 26 where there was almost no change in the bottom weld width of the specimens with focal position, Figure 28 indicates that when welding with filler wire, the bottom width also increased with increasing focal position. The increase in the root width at positive focal positions was due to enhanced interaction between laser beam and filler wire, in which the fraction of the reflected laser power was reduced and the absorption of the beam energy was maximised in terms of power density. The top weld width was less affected by this mechanism and so it increased with increasing magnitude of focal position, similar to previous observations. The top weld width was less than 4.0 mm at all focal positions, whereas, the bottom weld width was greater than 2.5 mm, at focal positions of +2 and +4 mm, so they failed the criterion in BS EN 4678. However, when taking into account the standard deviation in weld width measurement as indicated by the error bars in Figure 28, it may be possible to accept the bottom widths of these specimens.

73 Rw was above 0.6 for all specimens so a stable full penetration was achieved at all focal positions. The Rw was the largest at 0 mm, close to 1 and as it can be seen from Figure 29, the transverse cross-section of the weld at this focal position was very symmetrical with almost the same top and bottom weld widths. It was found that the Rw decreased with focal position which meant that the difference between the root and the face width increased with increasing magnitude of focal position. The underfill defect observed at 0 mm when autogenous welding, was absent when welding with filler wire. In fact, underfill was not observed at all focal positions. The large undercut observed at +4 mm when autogenous welding was also significantly reduced by welding with filler wire to a level close to the most stringent criteria of 0.05 mm in AWS D17.1. No undercut was observed at focal positions from 0 mm to -4 mm. Reinforcement defect which was not an issue when autogenous welding in the previous experiments, became a problem at 0 mm due to the high filler wire feed rate, with a height greater than the maximum of 0.55 mm in BS EN4678 but still below the criteria in AWS D17.1 and BS EN ISO 13919-2 of 0.99 and 0.65 mm, respectively. However, the problem with reinforcement was solved by defocusing the beam, by which the specimens passed all criteria.

Figure 29 clearly shows an improvement in the weld shape produced by welding with filler metal when compared to those in Figure 27 welded without, especially at focal positions of 0 and +4 mm. An hourglass shaped weld cross-section without crack and porosity was observed at all focal positions. However, overlap was observed at +2 mm. Overlap is a welding defect which forms at the weld toe or root caused by molten metal flowing on to the surface of the base metal without fusing to it. Often the cause of overlap is when the weld pool is large, inaccurate positioning of the workpiece or the presence of adherent surface oxides due to contamination which prevents the liquid metal from fusing with the base metal.

The asymmetrical weld cross-section at +2 mm indicated a possible misalignment of the wire feed position from the plane defined by the laser beam optical axis and the welding direction. Even a slight misalignment of 0.4 mm when welding with filler wire transverse to the welding direction can result in an asymmetrical weld [266],. This can be fixed by reducing the weld pool size, accurate positioning and alignment of the wire feed, laser beam and the workpiece, and finally, careful cleaning of the workpiece prior to welding. The optimum weld quality was therefore, obtained at focal positions of +4, -2 and -4 mm

74 a) 5 P = 4.9 kW, V = 3 m/min, Filler, Ar

2 Weld width (mm)width Weld

1 Top Bottom 0 -6 -4 -2 0 2 4 6 Focal position (mm) b) 1.2 0.50 Rw Reinforcement Undercut 1.0 0.40

0.8

0.30 , ,

w 0.6 R

0.20

0.4 Imperfections(mm) Reinforcement(mm)

0.10 0.2

0.0 0.00 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 Focal position (mm) Figure 28 a) Relationship between weld width and focal position at a laser power of 4.9 kW, a welding speed of 3.0 m/min, a filler metal feed rate of 5.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm -4 mm -2mm +0 mm +2 mm +4 mm Figure 29 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 4.9 kW, a welding speed of 3.0 m/min, a filler metal feed rate of 5.0 m/min and Ar

75 3.3.5 Effect of Filler Metal Feed Rate The weld metal composition in autogenous welding of AA 2024-T3 was initiailly identical to the base metal composition, so it was solidification cracking susceptible. Due to the higher equilibrium vapour pressure and lower boiling point of volatile alloying elements such as zinc (917°C) and magnesium (1091°C) compared to aluminium (2519°C), they were excessively and selectively vaporised during laser keyhole welding, and as a consequence, segregation of these elements and dissolution of strengthening precipitates [11] resulted in degradation of mechanical properties of the weld [267,268]. The vaporisation process involved transport of the alloying element from bulk to surface of the weld pool and then vaporisation at the liquid/vapour interface into the surround gas phases [42]. The reduced Mg content was found to be responsible for keyhole instability which induced macro porosity formation. One of the reasons is because the addition of elements such as zinc, silicon, copper and iron reduce the solubility of hydrogen in liquid aluminium due to strong bonding of Al atoms to these elements, whereas, strong attractive interactions between hydrogen and magnesium, lithium and titanium increase the hydrogen solubility [40]. The 2024 alloy also contained small quantities of silicon which induced the formation of non-strengthening coarse Mg2Si precipitates, and iron which induced the formation of Cu2FeAl7 phase, both of which reduced the fracture toughness of the weld significantly and depleted the solid solution strengthening Cu solutes in the matrix needed for aged hardening [242].

In order to reduce the cracking susceptibility of AA 2024-T3, it was necessary to adjust the chemical composition and the solidification microstructure of the weld metal to a less susceptible level. It was possible to reduce the crack sensitivity by introducing minor eutectic alloy such as Al-Si via the addition of a crack resistant filler metal of the right composition, usually from the 4xxx series alloys. More eutectics were provided by the filler wire to heal the cracks formed during the solidification of the weld pool. In addition, using filler wire also improved the process stability, reduced the tendency to porosity formation and led to wider volume of the weld pool by compensating for the loss of material due to vaporisation [232]. When filler metal was used, the weld metal composition was determined by the ratio between the amount of the melted base metal and the filler metal in the weld seam. The dilution ratio is usually around 20% for laser beam welding and due to the low fraction of filler metal in the weld, the eutectic alloy AA 4043 with 5% silicon was used as filler metal in this investigation. The major alloying element of the 4xxx series aluminium-silicon alloys is silicon in amounts up to a maximum of 12% in AA 4047, with a low melting point suitable as a filler material for welding AA 2024-T3 to eliminate solidification cracking and undercut, without causing brittleness in the resulting welded joint [35]. The high percentages of silicon in AA 4043 filler metal was important for the prevention of solidification cracking and it helped reduce the total

76 shrinkage during freezing. The low solidification temperature and narrow freezing range of around 5°C of the AA 4043 filler metal ensured that the base metal completely solidified prior to the weld and reached its maximum strength before becoming subjected by shrinkage stresses so that the duration in which liquid metal affected by this contraction was minimised during cooling. The filler metal was the last part of the weld to solidify. The AA 4043 filler wire also combined with the base metal to lower the percentage of MgSi2 in the weld and reduce its crack sensitivity. However, the addition of filler metal to the weld also led to a mismatching mechanical properties and reduced the yield and tensile strengths of the weld compared to the base metal. As well as that, in the case where the wire feed rate was too fast and not optimised, it resulted in weld pool instability and formation of welding defects. Therefore, the optimum values for wire feed rate was investigated to produce the best weld quality with reduced crack sensitivity.

(i) The effect of filler wire feed rate on the weldability and morphology of 3 mm thick AA 2024- T3 specimens was investigated. The filler wire feed rate was the only parameter which was changed while all the other parameters were kept constant. All specimens were bead on plate welds, welded using the optimised parameters in Section 3.3.4, with a laser power of 4.9 kW, a welding speed of 3.000 m/min, +4 mm defocus and helium gas shielding. The effect of varying the filler wire feed rate was studied from 0 m/min (autogenous) to 7.0 m/min.

Figure 30 shows that the top and the bottom weld width of the specimens increased with increasing filler wire feed rate. The change in the bottom width with feed rate was relatively small but quite large for the top width. The top weld width of specimens measured at all feed rates passed the criterion in BS EN 4678, of 4.0 mm, whereas the bottom weld width was above the maximum of 2.5 mm at all feed rates and therefore, failed the criterion. However, as the weld quality in general was good in terms of welding defects and Rw, and also because there were no weld quality acceptance criteria related to weld width specified in AWS D17.1 and BS EN ISO 13919-2, it was not possible to judge the specimens using the criteria on face and root weld widths in BS EN 4678 alone but also had to consider other factors as well. It was found that the magnesium content in the weld measured by EDX decreased with increasing feed rate from around 0.80% at 0 m/min to 0.65% at 7.0 m/min, which obviously increased the silicon content as well, from around 0.10% at 0 m/min to 0.88% at 7.0 m/min, illustrated in more details in Figure 32.

77 a) 5.0 2.0 Top Bottom Si Mg 4.0 1.5

3.0

1.0

2.0 Weight(%) Weld width Weldwidth (mm)

0.5 1.0

0.0 0.0 0 2 4 6 8 Filler wire feed rate (m/min) b) 1.2 0.50 Rw Reinforcement Undercut Underfill

1.0 0.40

0.8

0.30 ,

w 0.6 R

0.20 Imperfections Imperfections (mm)

Reinforcement Reinforcement (mm) 0.4

0.10 0.2

0.0 0.00 0.0 2.0 4.0 6.0 8.0 Filler wire feed rate (m/min) Figure 30 a) Relationship between weld width, magnesium and silicon content and filler metal feed rate at a laser power of 4.9 kW, a welding speed of 3.0 m/min, +4 mm defocus and with helium shielding gas, and b) the resultant weld width ratio, undercut, underfill and reinforcement

The Rw was above 0.6 at all feed rates so the processing stability for full penetration welding was high. A trend was observed where the Rw decreased with increasing feed rate, meaning that as mentioned above, the rate of change in the top width was greater than the bottom width with the feed rate. Underfill was observed when autogenous welding, above the maximum limit of 0.13 mm in AWS D17.1 and 0.15 in BS EN ISO 13919-2 but below 0.30 mm in BS EN 4678. The depth of underfill was reduced by more than half at 1.5 m/min which passed all criteria so welding with filler wire reduced the formation of underfill defects and even eliminated at higher feed rates. As expected the height of reinforcement or excess weld metal increased

78 with increasing feed rate and at 7.0 m/min, it was above the most stringent limit of 0.55 mm in BS EN 4678 but below that in BS EN ISO 13919-2 and AWS D17.1, of 0.65 and 0.99 mm, respectively. For the rest of the specimens, it was less than 0.55 mm. The depth of undercut was below the maximum limit of 0.15 mm in BS EN ISO 13919-2 and BS EN 4678 at all feed rates, but was above the 0.05 mm in AWS D17.1 at the two highest feed rates of 5.0 and 7.0 m/min. The problems with welding defects including surface porosity, reinforcement and undercut were found to be the most significant at 5.0 m/min and 7.0 m/min. It was possible that the feed rate was too high at these feed rates which supplied too much filler metal to the weld pool for the given laser power and welding speed. High feed rates produced instabilities, whereas, low feed rates did not sufficiently modify the chemical composition of the weld pool [33].

0 m/min 1.5 m/min 2.0 m/min 2.5 m/min

3.0 m/min 4.0 m/min 5.0 m/min 7.0 m/min Figure 31 Transverse sections of welds and weld top bead profiles produced with different filler metal feed rate at a laser power of 4.9 kW, a welding speed of 3.0 m/min, +4 mm defocus and with helium shielding gas

The weld shape was similar at all feed rates with an hourglass shape but with larger weld widths with increasing feed rate. The small underfill observed at 0 m/min was reduced at 1.5 m/min and completely removed at 2.0 m/min. The weld quality was good between 2.0 and 4.0 m/min but at 5.0 m/min, clustered pores with a maximum diameter of 0.21 mm. The criterion in AWS D17.1 for surface porosity specifies at least 8 times the size of larger adjacent imperfection and a smaller maximum pore size of 0.75 mm compared to 0.99 mm for subsurface pores. BS EN 4678 and BS EN ISO 13919-2 on the other hand, specify the same diameter of 0.99 mm for surface pores but with the distance between the individual pores in clustered porosity greater than ¼ of the material thickness. Although the size of surface pores

79 was small, they were too close to each other so were unacceptable. These surface pores were not observed at 7.0 m/min but instead large undercut defects on the top surface and excessive penetration on the bottom surface was found. Therefore, it was concluded that the best weld quality in terms of morphology was produced when the filler metal feed rate was in the optimum range of 1.5 to 4.0 m/min.

(ii) The effect of silicon addition on the crack sensitivity of AA 2024-T3 is shown in Figure 32. The weight content of magnesium and silicon in the weld was measured using energy dispersive X-ray spectroscopy on specimens welded with different filler metal feed rates. It was found that the magnesium level dropped whereas, the silicon level increased with increasing feed rate. As it can be seen from the crack sensitivity curves for aluminium in Figure 32, the crack sensitivity is the maximum when the Cu content is approximately 3%, Si is 1%, and Mg is 1.5%. AA 2024-T3 contained approximately 4.5% Cu which may initially have indicated that it has relatively low crack sensitivity. However, it also contained a small amount of Mg close to the critical level of 1.5% in the base metal, which increased the crack sensitivity by widening the coherence range, and depressing the solidus temperature but not the highest temperature of coherence [269]. Segregation of the low boiling point alloying elements such as Zn and Mg caused hot cracking at the grain boundaries due to the shrinkage strains during the solidification process. The presence of Si as well as Mg in the AA 2024-T3 base metal increased the risk of inducing coarse Mg2Si precipitates so the maximum content of Si in the 2024 alloy was required to be less than 0.7% [270].

According to Davis [259], the sensitivity decreases rapidly if the Si content exceeds 1.5%. The dilution of the weld pool with excess silicon by welding with the 4043 filler metal effectively reduced the percentage of Mg2Si in the weld by combining with the base metal. Also, the addition of silicon to the weld lowered the solidification temperature and decreased the total shrinkage during freezing as mentioned previously to prevent cracking. As a result, the peak of the solidification crack sensitivity curve for Al-Mg and Al-Mg2Si shifted away from the crack sensitive ranges. The silicon content in the welded specimens were detected to be less than the recommended 0.6% to avoid the crack sensitive range of Al-Si up to the feed rate of 3.0 m/min but above 0.6% at higher feed rates of 4.0, 5.0 and 7.0 m/min. Therefore, the solidification crack sensitivity was minimised by welding at a filler wire feed rate of 2.0 to 3.0 m/min.

80 a) b) 5 Vf = 0 m/min Vf = 1.5 m/min Vf = 2.0 m/min Vf = 2.5 m/min 4 Vf = 3.0 m/min Vf = 4.0 m/min Vf = 5.0 m/min Vf = 7.0 m/min 3

Weight(%) 2

0 Cu Mg Si Figure 32 a) Weight percentage (%) of main alloying elements in the weld as a function of filler metal feed rate obtained using energy dispersive X-ray spectroscopy (EDX) and b) aluminium crack sensitivity curves showing the effects of different alloy additions [246,271].

3.3.6 Effect of Shielding Gas Shielding gas was used in laser welding for several reasons. It protected the weld pool and the weld seam from the atmosphere to prevent oxidation and formation of welding defects such as porosity, and to improve the weld shape, size, quality and mechanical properties. Contaminations in the weld could reduce the effective laser power and cause keyhole instability. It also served to suppress and blow away the laser induced plasma and vapour which partially absorbed, scattered and attenuated the laser energy, so that the laser beam could efficiently reach the workpiece [89]. Part of the refracted laser beam by the interaction with vapour plume or plasma prevented the full power density in the incident laser beam from reaching the workpiece and influenced the keyhole geometry [272]. The plasma on the other hand, could also be considered beneficial as it assisted in coupling the beam energy to the weld pool. Two different mechanisms of heating the workpiece occurred during welding, one from significant absorption of beam energy from the keyhole and the other from the vapour plume above the weld, which released the previously absorbed laser energy into the weld near the surface to create a wider face width [273]. The observed weld shapes were therefore, controlled by the balance between these heat sources.

The characteristics of laser induced plasma is related to the wavelength of the laser source used. Welding aluminium alloys using CO2 laser with a larger wavelength of 10.6 μm induces strongly ionised plasma with temperatures over 16000 K [89] above the keyhole and partially absorb the CO2 laser radiation. On the other hand, near infrared shorter wavelength lasers such as Nd:YAG laser with 1.06 μm, disc laser with 1.03 μm and fibre laser with 1.07 μm wavelength show lower tendency to form such plasma and instead forms less ionised plume of metal vapour and shielding gas within the range of 3000-5000 K [89] which interacts

81 differently with shorter wavelength lasers. The inverse Bremsstrahlung absorption of laser beam in the vapour plume is proportional to the square of the laser wavelength so the absorption coefficient is around 100 times greater with the CO2 laser than the shorter wavelength fibre laser or Nd:YAG laser [274]. Therefore, the vapour plume is non- or only weakly ionised in these lasers while the vapour ionises to form a plasma in the CO2 laser.

Due to the apparent absence of a significant plasma, the plasma shielding effect is small so beam attenuation and scattering, and the widening of the beam intensity distribution was reduced [275–277]. Uspenskiy et al. [211] conducted spectral analysis of vapour plumed formed during 10kW high power fibre laser welding 6 mm Ti-VT-23 titanium alloy and detected around 3 cm high vapour plume forming on the surface of the workpiece. The vapour plume was only weakly ionised and so the absorbed radiation was negligible of less than 1%. Similarly, Kawahito et al. [275] found that the weakly ionised plume induced when fibre laser welding stainless steel at 10 kW level was not significant enough to influence the penetration depth. Gao et al. [172] on the other hand, studied the effect of laser power on the characteristics of fibre laser induced plume via emission spectroscopic analysis and found that a strong plasma shielding effect dominated by inverse bremsstrahlung absorption appeared when a laser power greater than 5 kW was used. The relatively smaller influence of ionisation on fibre laser induced plume was expected to enhance the welding process stability and lower the threshold power density for keyhole formation because of the enhanced Fresnel absorption and reduced beam attenuation and scattering. Katayama et al. [115] investigated the behaviour of laser induced plume or plasma and compared their interaction between CO2 and fibre laser for welding austenitic stainless steel and aluminium alloy. They found that shielding gas had a greater effect on CO2 lasers as compared to fibre lasers whereas, the type of shielding gas has less effect on Nd:YAG or fibre laser.

Helium and argon are the two most commonly used gases for laser welding aluminium alloys. They have different chemical and physical properties such as density, ionisation potential and thermal conductivity which affect their suitability as shielding gases. Argon gas has an ionisation potential of 15.8 eV, a thermal conductivity of 0.016 W/mK and a density of 1.661 kg/m3. Helium gas on the other hand, has a higher ionisation potential of 24.6 eV and thermal conductivity of 0.142 W/mK, but a lower density of 0.1664 kg/m3 [278]. Since the temperature and viscosity of plume is lower than plasma, the higher atomic weight and lower ionisation potential of argon may be more effective at plume control than helium for creating and then removing the vapour plume from the keyhole for the same shielding gas flow rate [272]. The lower ionisation potential of argon was not a critical issue because fibre laser was less sensitive to ionisation so argon would not ionise itself for the range of laser powers used in this investigation. Although helium can absorb greater amount of energy than argon before

82 becoming ionised, higher flow rates are required due to its lower density which comes at a greater cost. While both argon and helium gases are expensive, the cost of helium is almost 2.5 times greater than argon so the choice of shielding gases is also limited by their cost. However, the higher thermal conductivity of helium means that the energy input to the workpiece is greater and so a faster welding speed can be used to make up for the higher cost of helium. The shielding gas flow rate depends on type of shielding gas and welding parameters used such as laser power and welding speed. Higher flow rate is required for higher laser power or faster welding speeds.

(i) The influence of shielding gas composition on the weldability and morphology of 3 mm thick AA 2024-T3 specimens was investigated under two different welding conditions. The first set compared the difference between using argon and helium shielding gases as a function of welding speed and the second set compared the difference as a function of focal position, both of which were discussed previously in Section 3.3.3 and 3.3.4 using argon shield gas.

For the first set, a laser power of 4.9 kW and neither defocusing nor filler metal were used. As it can be seen from Figure 33, the difference in the top and the bottom weld widths between argon and helium was small. The weld widths were slightly greater when using helium and the difference gradually became larger with increasing welding speed so the extra heat potential of helium shielding gas was more effective at higher welding speeds. The bottom weld width of the specimen welded at 3.0 m/min when using helium was greater than the maximum root width of 2.5 mm specified in BS EN 4678 whereas, it was below 2.5 mm when using argon. Therefore, it was possible that the heat input was a little excessive at lower welding speeds when using helium. The same effect was observed with Rw, where the values were found to be very consistent with welding speed using helium, whereas, Rw increased with increasing welding speed using argon. The processing stability for full penetration welding was high for both argon and helium but it was less affected by a change in welding speed using helium. The level of reinforcement observed was acceptable according to the criteria in AWS D17.1, BS EN ISO 13919-2 and BS EN 4678 at all welding speeds for both argon and helium but still smaller in the specimens welded using helium. It was found that at the welding speed of 3.0 m/min, the depth of underfill was larger when using argon, close to the most stringent maximum limit of 0.13 mm in AWS D17.1. Although it was expected for the underfill to be greater due to the larger heat input when using helium, this was not the case because the extra energy from helium was used to produce much larger excessive penetration instead in the specimen welded using helium. The opposite trend was observed at higher welding speeds, where the undercut and underfill defects were larger when using helium than argon so in this case, more energy was transferred into the weld to produce these defects.

83 a) 5

2 Weld width Weldwidth (mm)

1 Top (He) Bot (He) Top (Ar) Bot (Ar) 0 2 3 4 5 6 Speed (m/min) b) 1.2 0.50 Rw (He) Undercut (He) Underfill (He) Reinforcement (He) Rw (Ar) Undercut (Ar) Underfill (Ar) Reinforcement (Ar) 1.0 0.40

0.8 0.30

w 0.6 R

0.20

0.4 Imperfections (mm)

0.10 0.2

0.0 0.00 2 3 4 5 6 Speed (m/min) Figure 33 a) Relationship between weld width and welding speed at a laser power of 4.9 kW, no defocus and with either argon or helium shielding gas, and b) the resultant Rw, undercut, underfill and reinforcement defects

The overall weld shape was an hourglass shape for both argon and helium. The weld profile improved using helium because the plume effect was reduced due to the higher ionisation potential of helium. The fusion boundaries were more parallel through thickness with lower reinforcement compared to the specimens welded using argon. The weld cross-section of the specimen welded at 5.0 m/min using argon was relatively narrower and as a result, a micro pore with a diameter of 0.064 mm was formed, but much smaller than the maximum pore size of 0.90 in BS EN 4678 and 0.99 mm in AWS D17.1 and BS EN ISO 13919-2. On the other hand, porosity was not observed at the same welding speed when using helium. The higher

84 heat produced in the helium shielded weld led to hotter weld pool and longer solidification time which allowed gas bubbles to escape and reduced the risk of gas entrapment. Therefore, it was concluded that the risk of porosity formation was reduced by using helium.

1.0 mm 1.0 mm 1.0 mm

3.0 m/min 4.0 m/min 5.0 m/min Ar

1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 3.0 m/min 4.0 m/min 5.0 m/min He Figure 34 Transverse sections of welds and weld top bead profiles produced with different welding speeds at a laser power of 4.9 kW, no defocus and with either argon or helium shielding gas

(ii) For the second set, a laser power of 4.9 kW, a welding speed of 3.0 m/min and no filler metal were used. As shown in Figure 35, the difference in the weld widths with focal position between argon and helium was negligible except at a focal position of +4 mm, where the bottom weld width of the specimen protected by helium was larger, above the maximum root width of 2.5 mm in BS EN 4678. The higher heat input associated with helium shielding gas was found to be more effective at larger positive focal positions because the effect of reduction in specific energy and lower average temperature in the weld pool due to positive defocusing was compensated. On the other hand, this effect was less significant at negative focal positions because the threshold power density for keyhole formation was lower than at positive focal positions. Due to the same reason above, larger undercut was produced using helium at focal positions from -4, -2 and +2 mm, all of which failed the criterion of 0.05 mm for undercut in AWS D17.1 and even failed the criteria of 0.15 mm in BS EN 4678 and BS EN ISO 13919- 2 at -2 mm, whereas, the specimens welded using argon all passed.

85 a) 5.0

4.0

3.0

2.0 Weld width (mm)width Weld

Top (He) 1.0 Bottom (He) Top (Ar) Bottom (Ar) 0.0 -6 -4 -2 0 2 4 6 Focal position (mm) b) 1.2 0.5 Rw (He) Rw (Ar) Undercut (He) Underfill (He) Reinforcement (He) Undercut (Ar) Underfill (Ar) Reinforcement (Ar) 1.0 0.4

0.8 0.3

w 0.6 R

0.2

0.4 Imperfections (mm)

0.1 0.2

0.0 0.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 Focal position (mm) Figure 35 a) Relationship between weld width and focal position at a laser power of 4.9 kW, a welding speed of 3.0 m/min and with either argon or helium shielding gas, and b) the resultant weld width ratio, undercut and reinforcement

In contrast, for the specimens welded at a +4 mm, the opposite happened where the depth of undercut and underfill was significant when using argon, whereas, these defects were non- existent when using helium, meaning that the higher heat input associated with helium was adequate in this case. Rw values were stable at all focal positions and there was almost no difference between argon and helium.

No cracking or porosity was observed at all focal positions for both argon and helium shielded specimens as shown in Figure 36. The weld shape was hourglass shaped for all specimens with small difference between helium and argon. However, the weld profile obtained using

86 argon was in general more uniform except at +4 mm, where large undercuts were formed both on the top and the bottom surfaces in the specimen welded using argon. On the other hand, the weld quality of the specimen welded using helium at +4 mm was good and better than at lower focal positions. Therefore, it was concluded that the use of helium was suitable at higher focal positions when the effective power density of the incident laser beam was lower.

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm -4.0 mm -2.0 mm 0 mm +2.0 mm +4.0 mm Ar

1.0 mm 1.0 mm 1.0 mm 1.0 mm

1.0 mm 1.0 mm 1.0 mm 1.0 mm 1.0 mm -4.0 mm -2.0 mm 0 mm +2.0 mm +4.0 mm He Figure 36 Transverse sections of welds and weld top bead profiles produced with different focal positions at a laser power of 4.9 kW, at a welding speed of 3.0 m/min and with either argon or helium shielding gas

3.3.7 Fillet Welding of AA 2024-T3 T-joints AA 2024-T3 sheets of 3.2 mm thickness were used both for the base plate and the stiffener of a T-joint. The welding parameters adapted for fillet welding of the T-joints were a laser power of 3.5 kW, welding speed of 1.8 m/min, AA 4043 filler metal of 1.0 mm in diameter at a feed rate of 5 m/min to prevent hot cracking, laser beam tilted by 30° (seam angle) from and focused on (no defocus) the surface of the workpiece and argon gas shielding at a flow rate of 20 l/min. The beam properties were the same as those given in Section 3.2.2. The T-joint was welded in two passes, one side at a time with the addition of filler wire.

Figure 37 shows that the weld cross-sections of fillet welded T-joints consisted mainly of the FZ, the HAZ and the BM. The morphology of the solidification structure was influenced by temperature gradient and solidification rate in each zone as discussed in Section 3.3. At the centre of the FZ where the solidification rate and the thermal gradient were lower, columnar to equiaxed transition (CET) was promoted and therefore, the formation of cellular dendritic

87 solidification structure with a fraction of fine equiaxed grains was identified. Close to the fusion boundary where the solidification rate and the thermal gradient were higher, columnar dendrites oriented along the direction of the thermal gradient were formed [249,279,280]. a) b)

2.0 mm 2.0 mm

c) d) 250 μm

250 μm

e) f)

50 μm 50 μm

The HAZ consisted of two different regions of a partially melted zone (PMZ) adjacent to the fusion line and an overaged zone next to the base metal. In the PMZ, coarsened grain boundaries occurred due to partial melting of the BM and dissolution of strengthening precipitates. In the overaged zone, coarsening of semi-coherent S´ (CuMgAl2) phase, and then transformation to incoherent stable S (CuMgAl2) phase occurred [242,244]. These precipitates became smaller with increasing distance from the fusion boundary [148,152]

88 Welding with filler metal produced crack free weld seams. The weld cross-sections were inspected from different locations along the length of the weld. As shown in Figure 37, in one of the weld cross-sections, welding defects were not observed. However, in the other weld cross-section, the weld seams contained both micro and macro porosities in the FZ. The size of the micro pore on the left produced from the second weld pass was 0.04 mm, whereas, the size of the macro pore on the right produced from the first weld pass was 0.38 mm. The micro pore was suspected to be caused by hydrogen gas entrapment based on its size. The macro pore on the other hand was more likely due to keyhole instability as it was much larger. Fortunately, the severity of porosity formation was much lower than the limits of 0.90 mm and 0.99 mm. Spatter and an underfill defect was found on the top surface of the same weld seam where macro porosity occurred. Spatter was caused by the ejection molten metal from the keyhole which landed on top of the keyhole to form a small crown of solidified metal. It also led to loss of metal which reduced the amount of molten metal available to fill the weld pool and resulted in the formation of underfill [115].

Reduction of micro porosity can be done by applying a better pre-weld cleaning immediately prior to welding, on the two sides of the T-joint where they meet and also the filler wire, to remove the surface oxide layer and other surface contaminants, as it minimises the time for oxide formation and moisture absorption. Reduction of welding defects related to keyhole instability such as macro porosity, underfill and spatter can be done by increasing penetration depth or the weld width. The ways in which the heat input supplied to the weld pool can be increase include for example, is by increasing the laser power, decreasing the welding speed and using helium shielding gas instead of argon.

3.4 Results and Discussion on Ti-6Al-4V Ti-6Al-4V welds were produced under various welding conditions of laser power, welding speed and focal position. All welds were autogenous bead on plate welds protected only by industrial grade argon shielding gas with 99.999% purity. The weld quality for each combination of welding parameters was assessed using the criteria shown in Table 7.

Table 7 Weld quality assessment criteria from Table 5 applied to 2 mm thick Ti-6Al-4V as defined by AWS D17.1, BS EN ISO 13919-2 and BS EN 4678 [19,231,240]

Face width Root width Porosity Undercut Reinforcement Standard Level (mm) (mm) (mm) (mm) (mm) AWS D17.1 Class A N/A N/A ≤ 0.66 ≤ 0.05 ≤ 0.66 BS EN ISO 13919-2 stringent B N/A N/A ≤ 0.60 ≤ 0.10 ≤ 0.50 BS EN 4678 Titanium alloys ≤ 3.20 ≤ 1.00 ≤ 0.40 ≤ 0.10 ≤ 0.40

In all the Ti-6Al-4V welded specimens investigated, weld underfill and root concavity were not present so every examined specimen passed the weld quality assessment criteria on underfill and root concavity. On the other hand, excess weld metal (top) or penetration (root) was

89 observed in all the specimens to a certain degree, but still much below the limits of 0.66, 0.50 and 0.40 mm defined in Table 7. The level of porosity detected was less than the size limit set for porosity of 0.66, 0.60 and 0.40 mm in Table 7 for all specimens. The largest pore observed had a diameter of 0.23 mm, which still met the requirements of Table 7. It was found that if any porosity was evident, it occurred in the lower half of the weld. The way in which this can be reduced is by improving the shielding gas arrangements especially the under-bead shielding gas.

No cracking defect was observed in all Ti-6Al-4V welded specimens as expected because titanium alloys are generally not considered susceptible to fusion zone solidification cracking and HAZ liquation cracking like aluminium alloys due to the absence of secondary phase dispersoids or precipitate particles, or impurities at the grain boundaries (70). Of more concern was embrittlement at elevated temperatures arising from contamination by hydrogen, nitrogen and oxygen absorption. An effective trailing shield was used to prevent this problem. Therefore, in this investigation the weld joint discontinuities were assessed mainly in terms of undercut, porosity, reinforcement, face and root widths and the ratio of root to face width, Rw.

Figure 38 shows the microstructure of the base metal (BM), the heat affected zone (HAZ) and the fusion zone (FZ) of the Ti-6Al-4V weld. The microstructure of the as-received mill-annealed Ti-6Al-4V at equilibrium and in room temperature consisted of mainly equiaxed α grains with some retained intergranular β phase as shown in Figure 38, where the bright regions are the equiaxed α and the dark regions are the intergranular β distributed at the elongated α grain boundaries. The equiaxed microstructure was obtained by mechanical working in the α - β phase field to transform lamellar α into equiaxed α in a transformed β matrix. The vanadium enriched β phase was only present in limited quantity at room temperature meaning that it made only a small contribution to strengthening because of its small proportion.

The microstructure of the HAZ between the FZ and the BM consisted of two sub-regions, one close to the FZ and the other close to the BM. The HAZ close to the FZ experienced higher temperatures during welding ranging from the β transus temperature of 980-995°C to the solidus temperature of 1605°C, whereas, the HAZ close to the BM was subjected to temperatures below the β transus temperature but high enough to affect its microstructure. The observed HAZ near the FZ consisted mostly of acicular martensitic phase α´ and a small amount of acicular α and primary α similar to the FZ. This kind of microstructure corresponds to a specimen quenched from a temperature below the β transus, similar to that observed by water quenching from 1100°C [185]. However, it was difficult to determine precisely the range of temperatures from which the cooling has occurred.

Figure 38 Microstructure of fibre laser welded Ti-6Al-4V in a) the base metal, b) the heat affected zone and c) the fusion zone at 200x and 500x magnifications etched using Kroll’s reagent

The microstructure of the HAZ close to the BM consisted of a mixture of the microstructural constituents in the FZ, a small amount of martensite α´, and the BM, mainly primary α and intergranular β phase [281]. The volume fraction of α´ in the HAZ decreased with increasing distance from the FZ boundary. The microstructure of the FZ contained fine acicular α´ with coarse columnar prior β grain boundaries which originated during weld solidification and grew opposite to the direction of heat flow from the partially melted β grains in the near HAZ to the weld centreline as shown in Figure 38.

The absence of grain boundary phase α along the prior β grain boundaries in the FZ indicates that the cooling rate after welding was greater than the critical cooling rate for diffusional

91 transformation of 410°C/s in Ti-6Al-4V [282]. The presence of grain boundary α along the prior β grains was the consequence of the cooling rate being close to the minimum limit necessary for diffusionless transformation. High cooling rates from temperatures above the martensite start temperature associated with the laser beam welding process promotes transformation of the body centre cubic (BCC), β into the hexagonal close packed (HCP) α´ microstructure.by a diffusionless transformation process [238]. The α´ is a metastable supersaturated α phase with an acicular morphology and its volume fraction increases with increasing cooling rate. Due to its formation by rapid nucleation and growth, it contains a higher dislocation density compared to the primary α grains with a smaller plate thickness. The smaller grain size and the higher dislocation density of α´ are responsible for the higher hardness of α´ than α [283].

Energy-dispersive X-ray spectroscopy was used for elemental analysis of welded specimens. Figure 39 shows the EDX spectrum of the FZ obtained from the scanning electron microscopes (SEM) and the analysed chemical composition of the parent material, the heat affected zone and the fusion zone. Minor differences in titanium, aluminium and vanadium contents were found in these three zones. Although small in difference, the percentage fraction of aluminium was in a decreasing order from the PM to the FZ. Aluminium has the lowest boiling point of 2519°C, while that of vanadium is 3407°C and titanium is 3287°C. As the maximum temperature during welding can reach above the boiling point of the material, and the loss of alloying elements by vaporisation was dependent on the boiling point of each element. As a result, the fraction of vanadium, with the highest boiling point, increased towards the weld, whereas, the other two decreased according to their vaporisation temperature. a) b) 100 89.94 89.78 89.43 90

60 PM HAZ 50 FZ

Weight(%) 40

10 5.43 5.30 5.12 4.63 4.92 5.45

0 Ti Al V

3.4.1 Effect of Laser Power The effect of laser power on the weldability and morphology of 2 mm thick Ti-6Al-4V specimens was investigated under three different welding conditions. The laser power was

92 the only parameter which was varied while all the other parameters were kept constant. Various laser powers were employed as listed in Appendix A.

(i) Figure 40 shows the effect of laser power at a low welding speed of 2.1 m/min on the weld transverse sections. It was found that different laser powers can result in different weld shapes. The two horizontal lines represent the requirements set by the welding quality acceptance criteria in BS EN 4678 in Table 7 for the face width and the root width. The face width should be below the blue line, and the root width should be above the red line and also below the blue line. It was found that the only specimen which failed these criteria was the one welded with the lowest power of 1.2 kW. Unlike all the other specimens where full penetration was obtained, this specimen showed incomplete penetration and according to Table 5, the specimen welded at 1.2 kW was not acceptable. Incomplete penetration resulted from too little heat input in proportion to the 1.2 kW laser power used. Both the top and bottom weld width increased with laser power to a maximum at around 2.4 kW. Any further increase in the laser power only resulted in a gradual decrease in both weld widths. The weld shape changed from an initial V shape at lower laser powers to an hourglass shape at higher laser powers due to increased magnitude of heat input transmitted to the material as shown in Figure 41.

The non-dimensional weld width ratio (Rw) for describing the weld shape was used to estimate the processing stability for a stable full penetration welding. All the welded specimens in Figure

40 showed Rw values above 0.4 shown by the upper horizontal red line, of around 0.7-0.8, except the specimen welded at 1.2 kW which gave a value of zero due to incomplete penetration. Therefore, for the rest of the specimens, the welding process was stable.

An undercut defect was observed in all specimens again except the specimen welded at 1.2 kW. The depth of undercut increased with laser power up to a maximum of around 0.15 mm. The criteria for undercut as shown by the two red horizontal lines where the lower one is for AWS D17.1 of 0.05 mm and the upper one is for both BS EN ISO 13919-2 and BS EN 4678 of 0.10mm, indicated that undercuts produced in the specimens welded using laser powers greater than 1.5 kW were all above the maximum limit specified in AWS D17.1, whereas, the specimens welded with laser powers up to 1.8 kW were accepted according to BS EN ISO 13919-2 and BS EN 4678. Therefore, it was determined that at a low welding speed of 2.1 m/min used, the optimum processing parameter window was very narrow in terms of undercut since only the specimens welded at 1.5 kW and 1.8 kW passed the criteria. The main reason for the increasing depth of undercut is excessive power or heat input, which causes the evaporation and expulsion of the molten materials from the sides of the welded joint into the weld, leaving a drain-like impression along the length of the weld [185,190,191].

93 a) 3.5

3.0

2.5

2.0

1.5 Weld width (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 Laser power (kW) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 Laser power (kW) Figure 40 a) Relationship between weld width and laser power at a welding speed of 2.1 m/min with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm 2 mm 2 mm

1.2 kW 1.5 kW 1.8 kW 2.1 kW 2.4 kW

2 mm 2 mm 2 mm 2 mm

2.7 kW 3.0 kW 3.3 kW 3.6 kW Figure 41 Transverse sections of welds produced with different laser powers at a welding speed of 2.1 m/min with +4 mm defocus and argon shielding gas

94 The influence of laser power on weld reinforcement was not significant in the analysed specimens including the specimen welded at 1.2 kW. The amount of reinforcement was similar for all specimens in the range between 0.10 and 0.20 mm, which was well below the limit of 0.40 mm specified in the most stringent criteria AWS D17.1 for reinforcement. Porosity was observed in the specimens welded with the lowest three laser powers of 1.2, 1.5 and 1.8 kW as shown in Figure 41, those of which either lack of penetration or with relatively narrower root width. It is well known that welds with incomplete penetration tend to have higher amount of porosity [284]. According to a statistical analysis of defects in laser beam welded Ti-6Al-4V by Murav’ev [285], it was found that pores account for 43-56% of the total number of defects. The formation of porosity depends on the heat input. Increasing the welding power also increases the heat input and reduces the cooling rate to allow gases to escape from the weld pool and thereby reduce the risk of porosity. Such a trend was observed by Fan et al. who found high levels of porosity produced in the FZ when laser welding 0.7 mm thick Ti-6Al-4V with a low laser power of 0.6 kW and a welding speed of 4.0 m/min [206] and increasing the heat input reduced the probability of pores being retained in the FZ because the liquid metal solidification rate was delayed and so there was more time for bubbles to escape the weld pool. Pores were found along the FZ centreline in the lower half of the weld with a pore diameter of 0.10, 0.05 and 0.13 mm, respectively. The size of these pores were smaller than the maximum tolerable size of 0.40 mm specified in BS EN 4678, 0.60 mm in BS EN ISO 13919-2 and 0.66 mm in AWS D17.1. They were all spherical in shape and since there is no element in Ti-6Al-4V with a low boiling point, porosity was caused by the presence of gases such as hydrogen, nitrogen, oxygen or argon shielding gas from the environment and contamination of the workpiece.

Hydrogen solubility in Ti-6Al-4V is a function of temperature and as the temperature drops during solidification, the solubility of hydrogen increases, which promotes the rejection of hydrogen into the FZ and pores are formed when the absorbed hydrogen cannot escape the molten weld pool before solidification [286,287]. The small size of pores measured indicated that they were probably formed through diffusion of hydrogen [288]. Although it was previously reported that welds with an hourglass shaped cross-section have a greater tendency to entrap gases than nail head shaped welds due to the downward sweeping solidification fronts in the root of hourglass shaped welds [72], it was discovered in this investigation that low internal porosity content was observed in hourglass shaped welds. On the other hand, the V shaped welds at lower laser powers showed increasing propensity for porosity formation since the escape of gas bubbles was only possible via the top surface and also the solidification rate increased as the weld width decreased, whereas, in fully penetrated welds, more effective escape was allowed via both the top and bottom surfaces [268]. Therefore, the best weld quality when welding at 2.1 m/min was produced at a laser power of 1.8 kW.

95 (ii) Figure 42 shows the effect of laser power when a faster welding speed of 3.9 m/min was used. The number of partially penetrated specimens increased when the welding speed increased from 2.1 m/min to 3.9 m/min. These included the specimens welded at 1.8 and 2.1 kW, which were previously fully penetrated but only partially penetrated at 3.9 m/min as a result of reduced heat input at a higher speed. Analysis of the weld widths showed that the top width of all specimens satisfied the 3.2 mm limit in AWS D17.1 but the root width of the specimens welded up to 2.7 kW were less than the minimum of 1.0 mm. A trend was observed where the top and bottom weld widths increased with laser power up to 3.3 kW and then levelled off beyond that. Compared to the weld produced at 2.1 m/min, the weld widths were around 1.0 mm narrower at 3.9 m/min for the same laser powers. The weld shape in Figure 43 was also narrower with a nail head shape at lower laser powers up to 3.0 kW and then switched to an hourglass shape above 3.3 kW. The Rw corresponded to the change in the weld shape where it increased from 0 to around 0.6 with laser power up to 3.0 kW with a nail head shaped weld which then levelled off with an hourglass shaped weld. The Rw of the specimens welded at 1.8, 2.1 and 2.4 kW were below 0.4 as shown by their nail head weld shape. On the other hand, the depth of undercut formed in these three specimens passed the AWS D17.1 criterion of 0.05 mm. The level of undercut observed in the specimens welded at 2.1 m/min was greater than the less stringent BS EN ISO 13919-2 and BS EN 4678 criteria of 0.10 mm at laser powers above 2.1 kW. On the other hand, at 3.9 m/min, the undercut was below 0.10 mm over all laser power used. In addition, all welds confidently passed the criteria on reinforcement of all standards without any obvious trend in relation to laser power.

Porosity was only detected in the specimen welded at 2.1 kW as shown in Figure 43 with a diameter of 0.10 mm at the root of the weld, less than the most stringent size of 0.40 mm in BS EN 4678. The cause of this porosity was suspected to be hydrogen gas entrapment. From the results above, it was determined that a wider operating window was obtained when the welding speed was increased to 3.9 m/min, with the welds welded at laser powers above 3.0 kW passing all the criteria in BS EN ISO 13919-2 and BS EN 4678, and also all except undercut in AWS D17.1.

96 a) 3.5

3.0

2.5

2.0

1.5 Weld width Weldwidth (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 Laser power (kW) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 Laser power (kW) Figure 42 a) Relationship between weld width and laser power at a welding speed of 3.9 m/min with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm 2 mm 2 mm

1.8 kW 2.1 kW 2.4 kW 2.7 kW 3.0 kW

2 mm 2 mm 2 mm 2 mm 2 mm

3.3 kW 3.6 kW 3.9 kW 4.2 kW 4.5 kW Figure 43 Transverse sections of welds produced with different laser powers at a welding speed of 3.9 m/min with +4 mm defocus and argon shielding gas

97 (iii) The effect of laser power at a fast welding speed of 6.0 m/min was also investigated as shown in Figure 44. A similar trend to welding at 3.9 m/min was observed but with smaller top and bottom weld widths by around 0.5 mm. Partial penetration was observed at 2.1 and 2.4 kW, which with 2.1 m/min produced fully penetrated hourglass shaped welds and with 3.9 m/min produced fully penetrated nail head shaped weld only at 2.4 kW.

The top weld width of all specimens were below the maximum limit of 3.2 mm, whereas, the bottom width of specimens welded at 2.1 to 3.0 kW were less than the minimum of 1.0 mm required. The bottom width of specimens welded at above 3.4 kW were also very close to 1.0 mm. The weld shape was V shaped at 1.8 kW which changed to nail head shaped at 2.1 kW and remained the same all the way up to the maximum laser power of 4.5 kW investigated, which indicated a low heat input to the workpiece at all power levels.

The Rw showed a similar trend as discussed above for 3.9 m/min, the Rw increased with laser power up to 3.4 kW and then levelled off to a value of around 0.6 above 3.4 kW. The Rw of welds produced at 2.1, 2.4 and 2.7 kW were below 0.4 meaning that the welding process was unstable below 2.7 kW. Undercut measured in the specimens at 6.0 m/min were all very close to 0.05 mm from AWS D17.1. The specimen which satisfied all the criteria on root width, Rw and undercut was the one welded at 3.6 kW. Reinforcement measured in this specimen was the greatest but still well below the minimum limit of 0.4 mm in BS EN 4678.

Formation of porosity was observed in the specimens welded at 2.4, 2.7 and 3.0 kW, all of which had a root width which failed the criterion in BS EN 4678. The pores were found along the weld centreline with a diameter of 0.08, 0.14 and 0.08 mm for 2.4, 2.7 and 3.0 kW, respectively, all below the most stringent size of 0.40 mm in BS EN 4678. Therefore, it was possible to obtain good quality welds at a laser power greater than 3.4 kW when welding at 6.0 m/min, which showed level of imperfections acceptable by all standards.

98 a) 3.5

3.0

2.5

2.0

1.5 Weld width Weldwidth (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 Laser power (kW) b) 1.2 0.30 Width ratio Undercut 1.0 Reinforcement 0.25

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 Laser power (kW) Figure 44 a) Relationship between weld width and laser power at a welding speed of 6.0 m/min with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm 2 mm 2 mm

2.1 kW 2.4 kW 2.7 kW 3.0 kW 3.4 kW

2 mm 2 mm 2 mm 2 mm

3.6 kW 3.9 kW 4.2 kW 4.5 kW Figure 45 Transverse sections of welds produced with different laser powers at a welding speed of 6.0 m/min with +4 mm defocus and argon shielding gas

99 3.4.2 Effect of Welding Speed The effect of welding speed on the weldability and morphology of 2 mm thick Ti-6Al-4V specimens was investigated under four different welding conditions. The welding speed was varied while the other parameters were kept constant. Various welding speeds were employed as listed in Appendix A.

(i) Fully penetrated welds were obtained at all welding speeds from 1.8 to 4.2 m/min, but the root width of the specimens welded at higher welding speeds were very narrow. Lack of penetration in the specimens welded at higher speeds arose from the low heat input as a result of the fast welding speeds so a careful balance between heat input and welding speed was necessary. Figure 46 shows that the top and bottom weld widths were inversely proportional to the welding speed so increasing the welding speed rendered narrower weld widths. Both the top and bottom weld widths were relatively wide at low welding speeds but the change in the root width was more significant than the top width with increasing welding speed. The top width of all specimens were below the maximum limit for face width in BS EN 4678 whereas, the bottom width of specimens welded at 3.6 and 4.2 m/min failed the criterion on minimum root width. It was found that welding speed influenced the weld pool shape by affecting the liquid metal flow during welding, and so produced different weld geometries in Figure 47.

A sharp decrease in the root width resulted in a change of weld shape from hourglass shape at low welding speeds to nail head shape above 3.0 m/min. The Rw for the specimens as shown in Figure 46 decreases quickly with increasing welding speed, indicating that the rate of change of the top width was relatively small compared to the bottom width. The specimens welded at a welding speed greater than 3.6 m/min gave Rw values smaller than 0.4 so the welding process was considered unstable at these welding speeds.

Undercut defects were observed both on the top and the root surfaces. The depth of undercut measured in all specimens were very small, with values less than the maximum tolerable depth of undercut specified in the most stringent AWS D17.1 of 0.05 mm. The undercut depth initially increased with welding speed up to 3.0 m/min, where it reached the maximum and then started to decrease with further increase in welding speed. According to Figure 46, at higher welding speeds or in nail head shaped welds, less undercut was observed due to low heat input and less expulsion and evaporation of molten materials. Similarly, the height of reinforcement decreased with increasing welding speed, although not significant, from around 0.20 mm at 1.8 m/min to 0.15 mm at 4.2 m/min. As mentioned previously, all specimens showed reinforcement well below the 0.40-0.66 mm limits in the standards, which would otherwise increase non-value added costs. It was therefore, concluded that fast welding speed is preferred to minimise undercut and reinforcement, whereas, slow welding is preferred to obtain a stable full penetration in terms of Rw.

100 a) 3.5

3.0

2.5

2.0

1.5 Weld width (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 6 7 8 Speed (m/min) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 6 7 8 Speed (m/min) Figure 46 a) Relationship between weld width and welding speed at a laser power of 2.0 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm 2 mm 2 mm

1.8 m/min 2.4 m/min 3.0 m/min 3.6 m/min 4.2 m/min Figure 47 Transverse sections of welds produced with different welding speeds at a laser power of 2.0 kW with +4 mm defocus and argon shielding gas

101 The amount of porosity as mentioned previously depended on the rate of weld pool solidification. At lower welding speeds, the heat input increased which effectively reduced the cooling rate in the weld pool during solidification, allowing gas bubbles to escape. The weld pool remains in liquid state for a longer period at lower welding speeds, which favours the nucleation, growth and finally the escape of the gas bubbles. Keyhole instability and collapse may also contribute to the formation of porosity at lower welding speeds [191]. At very high welding speeds, the solidification of the weld pool is too fast that there is insufficient time for the growth of pores so the risk of porosity formation is reduced. The chance of observing porosity is the greatest at intermediate speeds, where there is enough time for pores to grow but also for the pool to solidify before the escape [286]. In fact, the largest pore was observed in the specimen welded at the intermediate welding speed of 3.0 m/min, with a diameter of 0.23 mm, which was still less than the maximum pore size of 0.40 mm tolerated in BS EN 4678. As shown in Figure 47, porosity was observed over the entire range of welding speeds used, all along the weld centreline at lower half of the welds, including the specimens welded at 1.8 m/min and 4.2 m/min, with a pore diameter of 0.12 and 0.10 mm, both of which passed the criterion above.

(ii) Figure 48 shows the relationship between welding speed at a fixed laser power of 2.5 kW, and the weld widths. A very wide face and root widths were obtained at 1.8 m/min, with the top width exceeding the limit of 3.2 mm in BS EN 4678 and the bottom width also being very close to the same upper limit of 3.2 mm. Compared to the specimens welded at 2.0 kW, the specimens welded at 2.5 kW showed a steeper decrease in the top width with welding speed, whereas, the bot width decreased at a similar rate as before. In general, both the top and bottom weld widths were wider at 2.5 kW than at 2.0 kW for the corresponding speeds. The specimen welded at 1.8 m/min did not pass the criterion for the face width and the specimens welded at 4.2 and 4.8 m/min did not pass the criterion for the root width. The transverse section of the specimen welded at 1.8 m/min in Figure 49 has a noticeably wide face and root weld widths with almost rectangular weld shape. The specimen welded at a faster welding speed of 2.4 m/min displayed an hourglass shape, which changed into a nail head shape at 4.2 m/min where the root width became too narrow according to the 1.0 mm minimum root width specified in BS EN 4678. The trend in Rw was the same as that for the specimens welded at 2.0 kW, where it dropped below the limit of 0.4 at the same welding speed of 4.2 m/min. The specimen welded at 1.8 m/min produced undercut with a depth greater than the maximum undercut depth of 0.10 mm cited in the less stringent BS EN ISO 13919-2 and BS EN 4678. The specimens welded at 2.4 and 3.0 m/min passed the criterion of BS EN ISO 13919-2 and BS EN 4678 but failed that of AWS D17.1 which is 0.05 mm.

102 a) 3.5

3.0

2.5

2.0

1.5 Weld width Weldwidth (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 6 7 8 Speed (m/min) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 6 7 8 Speed (m/min) Figure 48 a) Relationship between weld width and welding speed at a laser power of 2.5 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm

1.8 m/min 2.4 m/min 3.0 m/min

2 mm 2 mm 2 mm

3.6 m/min 4.2 m/min 4.8 m/min Figure 49 Transverse sections of welds produced with different welding speeds at a laser power of 2.5 kW with +4 mm defocus and argon shielding gas

103 The specimens welded at higher welding speeds of 3.6, 4.2 and 4.8 m/min, satisfied all criteria with undercut depths below 0.05 mm. The height of reinforcement in specimens were all below 0.4 mm, with a maximum of around 0.25 mm, so they were all acceptable. The reinforcement decreased with increasing welding speed. As mentioned above, increasing the welding speed had a positive influence on reducing the extent of imperfections in the welds, whereas, a sufficiently fast welding speed was necessary to achieve a stable full penetration.

Porosity was observed in every specimen for the welding speeds investigated at the given power of 2.5 kW. The pores were all round in shape and their diameters were 0.04, 0.08, 0.04, 0.05, 0.14 and 0.10 mm for the welding speed of 1.8, 2.4, 3.0, 3.6, 4.2 and 4.8 m/min, respectively. The pores were the largest in the specimens which were welded at the two fastest welding speeds of 4.2 and 4.8 m/min and with nail head shaped welds. Although it was discussed previously that porosity decreased with increasing welding speed, the narrow root width of the specimens welded at higher speeds resulted in a greater probability of retaining porosity than in wider welds at lower speeds [260,285,286], because the escape of gas porosity before solidification became more difficult with a smaller root width and as a result, was only accessible via the top surface [268]. As determined by the low Rw at 4.2 and 4.8 m/min, it was believed that keyhole instability and collapse at very high welding speeds contributed to an increase in the size of pores. Nevertheless, the size of all pores were confidently within the safe limit of 0.40 mm in BS EN 4678.

(iii) A similar trend to the one given in Figure 46 where increasing the laser power beyond 2.4 kW only led to a decrease in the weld width for a given welding speed, was also observed in Figure 50. The weld widths of specimens for different welding speeds, welded at 3.0 kW, did not increase compared to those welded at 2.5 kW. Instead, the opposite was observed where the weld widths of the specimens welded at 3.0 kW decreased compared to those at 2.5 kW. Also, the rate at which the top and the bottom weld widths decreases with welding speed became lower. When a 3.0 kW laser power was used, the operating window for welding speed became wider as shown by the weld widths of the specimens welded at welding speeds of 4.2, 4.8 and 5.4 m/mins, those of which at previously investigated lower laser powers failed the criterion on the minimum root width of 1.0 mm accepted in BS EN 4678.

In the specimens welded at 3.0 kW, the top weld width was acceptable over the entire welding speed range used. The bottom weld width of all specimens except for the one welded at the fastest welding speed of 6.0 m/min, passed the criterion above. For this combination of welding speeds and laser power, all specimens had Rw values above 0.4 meaning that a stable full penetration was obtained in all of them.

104 a) 3.5

3.0

2.5

2.0

1.5 Weld width Weldwidth (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 6 7 8 Speed (m/min) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 6 7 8 Speed (m/min) Figure 50 a) Relationship between weld width and welding speed at a laser power of 3.0 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm 2 mm

1.8 m/min 2.4 m/min 3.0 m/min 3.6 m/min

2 mm 2 mm 2 mm 2 mm

4.2 m/min 4.8 m/min 5.4 m/min 6.0 m/min Figure 51 Transverse sections of welds produced with different welding speeds at a laser power of 3.0 kW with +4 mm defocus and argon shielding gas

105 The weld shape was an hourglass shape at low welding speeds up to around 3.6 m/min and then gradually changed to a nail head shape but still with some features resembling the hourglass shape. There was only a small variation in the depth of undercuts produced in these specimens, all of them passing the limit of 0.10 mm set in BS EN ISO 13919-2 and BS EN 4678, except for the specimen welded at 2.4 m/min which still passed but was very close to the limit. On the other hand, none of the specimens passed the more stringent 0.05 mm criterion in AWS D17.1 but not by a significant amount. The influence of welding speed on reinforcement was obvious for these specimens with a steady reduction in the height of the reinforcement with increasing welding speed. Overall, the amount of reinforcement in these specimens were well acceptable by all criteria below 0.4 mm.

Formation of porosity was not observed in any specimen, for all welding speeds in the range between 1.8 and 6.0 m/min, at a given laser power of 3.0 kW. Under these combinations of welding parameters, welds free of porosity were produced. Except for the specimen weld at a welding speed of 6.0 m/min, which had too narrow root width, any speed below 6.0 m/min can therefore, be used to produce a good quality weld, meaning that a wide operating window was obtained.

(iv) A further reduction in the overall weld widths were observed in the specimens as shown in Figure 52 for different welding speeds at a given laser power of 3.5 kW, compared to that of 2.5 and 3.0 kW. The change in face and root weld widths for these specimens with welding speed was not significant. The specimen welded at the lowest welding speed of 2.4 m/min had the widest top and bottom weld widths. For the other specimens, the top and the bottom weld widths only varied by around 0.2 mm from 3.6 m/min to 7.2 m/min, with values close to 2.0 mm and 1.0 mm, respectively. The bottom widths of the specimens welded greater than 2.4 m/min were very close to the minimum limit of 1.0 mm cited in BS EN 4678. The top widths, on the other hand, were less than the maximum limit of 3.2 mm by safely more than 1 mm.

The weld shape at 2.4 m/min was an hourglass shape and then nail head shape above

3.6 m/min with considerably narrow root widths. The Rw shown in Figure 53 was found to decrease with increasing welding speed. The Rw was the largest of around 0.8 at 2.4 m/min as illustrated by its hourglass weld shape, and then decreased to around 0.6 for the specimens welded at higher welding speeds as shown by their nail head shaped welds. As the Rw of all specimens were above 0.4, the welding process produced a stable full penetration for all welding speeds investigated. An undercut was observed, but its dependency on welding speed was found to be very small. For all the specimens welded at 3.5 kW, the undercut remained fairly constant regardless of welding speed, with values within the less stringent 0.10 mm and the more stringent 0.05 mm limits. Reinforcement or excessive root penetration was observed in the specimen welded at 2.4 m/min with a height of around 0.25 mm at bottom.

106 a) 3.5

3.0

2.5

2.0

1.5 Weld width Weldwidth (mm) 1.0

0.5 Top Bot 0.0 0 1 2 3 4 5 6 7 8 Speed (m/min) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

0.6 0.15 Rw

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 0 1 2 3 4 5 6 7 8 Speed (m/min) Figure 52 a) Relationship between weld width and welding speed at a laser power of 3.5 kW with +4 mm defocus and argon shielding gas and b) the resultant weld width ratio, undercut and reinforcement

2 mm 2 mm 2 mm 2 mm 2 mm

2.4 m/min 3.6 m/min 4.8 m/min 6 m/min 7.2 m/min Figure 53 Transverse sections of welds produced with different welding speeds at a laser power of 3.5 kW with +4 mm defocus and argon shielding gas

107 Reinforcement decreased with increasing welding speed and its effect on weld quality was insignificant and well below the 0.4 mm limit. Porosity was only observed in the specimen welded at the fastest welding speed of 7.2 m/min. However, the size of the pore was only 0.04 mm which was much smaller than the maximum size of 0.4 mm in BS EN 4678. Interestingly, for this particular specimen, a spatter was found at the bottom surface near the fusion boundary but there were no welding quality acceptance criteria on spatter according to AWS D17.1, BS EN ISO 13919-2 and BS EN 4678, except in BS EN ISO 13919-2 which states that the acceptance of spatter depends on applications. However, the effect of spatter found in this specimen on weld quality or performance was considered to be small.

3.4.3 Effect of Focal Position The effect of focal position on the weldability and morphology of 2 mm thick Ti-6Al-4V specimens was investigated. The focal position was the only parameter which was changed while all the other parameters were kept constant. Various focal positions were employed as listed in Appendix A. Focal position was defined as the distance relative to the top surface of the workpiece, where a value of zero corresponds to a beam focused on the top surface of the workpiece, positive values above and negative values below the top surface of the workpiece.

Figure 54 shows the effect of varying focal position on the weld width. The location of the focal position determines the intensity of the beam on the surface of the workpiece. While the focal position does not significantly affect the depth of penetration in thin materials, often it is located within the material when welding relatively thick materials in order to optimise the depth of penetration. The laser beam energy density is the maximum at the focal point so by focusing the beam either above or below the workpiece, the beam intensity can be reduced, which leads to a wider weld pool and larger keyhole which can be beneficial for processing stability. On the other hand, it may also be desired to have a small focal spot as it means that there is less material involved during welding. Positive defocusing reduces the irradiance on the surface but the keyhole is inherently more stable. Although it was observed by Pastor et al. [255] that positive defocusing leads to more plasma which defocuses the beam, this is not a serious problem with fibre laser as the plasma effect is very small with fibre laser due to its wavelength. Negative defocusing leads to an increased laser beam coupling to the material and increases the irradiance inside the weld pool [142] so the welding speed can be increased than without defocusing [82]. It was therefore, important to keep the focal distance within the appropriate range to maintain the required power density to form a keyhole. The optimum focal position would result in the maximum Rw, suitable face and root widths and the minimum number of imperfections.

108 Full penetration was achieved in all welds regardless of the focal position used as shown in Figure 54. The top and the bottom weld widths were the minimum when the beam was focused on the surface of the workpiece, 0 mm. The general trend was that the weld widths increased with both positive and negative focal positions. As the threshold power density for keyhole formation is lower at negative focal position due to improved laser and material interaction [255], the weld widths at negative focal positions were larger than that for the same extent of positive focal positions at a given laser power of 2.5 kW and welding speed of 3.0 m/min. The BS EN 4678 criteria on face and root widths of 3.2 mm and 1.0 mm, respectively were satisfied at all focal positions investigated. a) 3.5

3.0

2.5

2.0

1.5 Weld width Weldwidth (mm) 1.0

0.5 Top Bot 0.0 -6 -4 -2 0 2 4 6 Focal position (mm) b) 1.2 0.30 Width ratio Undercut 1.0 0.25 Reinforcement

0.8 0.20

w 0.6 0.15 R

0.4 0.10 Imperfections(mm)

0.2 0.05

0.0 0.00 -6 -4 -2 0 2 4 6 Focal position (mm) Figure 54 a) Relationship between weld width and focal position at a laser power of 2.5 kW, a welding speed of 3.0 m/min and with argon shielding gas, and b) the resultant weld width ratio, undercut and reinforcement

The Rw was also satisfied at all focal positions, all above the 0.4 mm limit. It was found that the Rw was greater at smaller focal positions of ±2 mm and 0 mm, with the largest value for +2 mm, and smaller at larger focal positions of ±4 and ±5.5 mm. In terms of the weld shape, for smaller focal positions, the weld shape was an hourglass shape, whereas, at larger focal positions, the weld shape was a V shape. Undercut in general, was fairly small at all focal

109 positions because the laser power and the welding speed used were already optimum values and also partly because the heat input was not affected by the focal position but only the laser power density changed. The depth of undercut was all below or very close to the 0.10 mm limit in BS EN 4678 and BS EN ISO 13919-2. The specimens welded at +2 and -4 mm focal positions showed the largest undercut which marginally passed the criteria in BS EN 4678 and BS EN ISO 13919-2. The specimen welded at +5.5 mm focal position produced the smallest undercut which passed the 0.50 mm maximum depth of undercut in AWS D17.1. The size of undercut measured in the specimens welded at 0 and +4 mm focal positions was also reasonably small. There was only a small change in the height of reinforcement with focal position except the specimen welded at a focal position of 5.5 mm which exhibited excessive penetration as previously explained by the laser-material interaction for negative defocusing.

As it can be seen from Figure 55, porosity was formed at -4.0 and +5.5 mm, with a diameter of 0.04 and 0.03 mm, respectively. The specimens with porosity had small Rw values. The best weld profile was obtained at +4 mm where a stable full penetration was achieved with a sufficiently wide root width, small undercut and reinforcement, and no porosity. The specimen welded at -2 mm also produced a good weld quality comparable to that at +4 mm.

Figure 55 Transverse sections of welds produced with different focal positions at a laser power of 2.5 kW, a welding speed of 3.0 m/min and with argon shielding gas

3.4.4 Fillet Welding of Ti-6Al-4V T-joints Ti-6Al-4V sheets of 2 mm thickness were used both for the base plate and the stiffener of a T-joint, fillet welded using a laser power of 3.1 kW, a welding speed of 6.0 m/min, filler metal of the same Ti-6Al-4V with a 1.0 mm diameter at a feed rate of 3 m/min, laser beam tilted by 30° (seam angle) from and focused on (no defocus) the surface of the workpiece and argon gas shielding at a flow rate of 20 l/min. The beam properties were the same as those given in

110 Section 3.2.2. The T-joint was welded in two passes, one side at a time with a filler wire. Micrographs of T-joint weld cross-sections in Figure 56 show that a good quality weld seam was produced which completely fused the stiffener and the base plate, with a single common molten pool that is almost symmetrical relative to the stiffener centreline. The T-joint consisted of four distinct zones having different microstructures, including fusion zone, heat affected zones, one close to the fusion zone and the other close to the base metal, and base metal.

As discussed above in Section 3.4, the fusion zone had a fine needle like martensitic structure with coarse prior columnar β grain boundaries. The near HAZ consisted of mostly of acicular martensitic phase α´ and a small amount of acicular α and primary α similar to the FZ and its

111 width was only around 0.1 mm wide. The far HAZ showed a small amount of martensite α´, mostly primary α and intergranular β phase, with a slightly wider width of around 0.3 mm. The BM comprised equiaxed α grains with some retained intergranular β phase.

The weld seam was crack free but small micro pores of 0.048 and 0.034 mm diameters were found at the boundary between the FZ and the HAZ due to hydrogen gas entrapment rather than keyhole instability, based on their size. For the T-joint, the escape of gas bubbles was only possible from the top surface. The level of weld metal porosity achieved was lower than that defined for the stringent quality class in BS EN 4678 of 0.40 mm. The possible methods of reducing the amount of porosity in the T-joint would be to apply pre-weld cleaning immediately prior to welding, on the two sides of the T-joint where they meet and also the filler wire, to remove the surface oxide layer and other surface contaminants, as it minimises the time for oxide formation and moisture absorption.

3.5 Conclusions

3.5.1 AA 2024-T3 The fusion zone consisted of fine equiaxed and columnar dendrites. The heat affected zone consisted of a partially melted zone close to the fusion zone where dissolution of the strengthening precipitates occurred and also an overaged zone next to the base metal where over-ageing by coarsening of the strengthening precipitates, semi-coherent S´ (CuMgAl2) phase, and then a transformation to the incoherent stable S (CuMgAl2) phase occurred. T

Excessive undercutting was observed at the highest power density. Solidification cracking occurred at all power densities so the addition of filler metal was required to reduce crack sensitivity. Insufficient laser power resulted in lack of fusion. Increasing the laser power improved the weld quality significantly. Defocusing the laser beam deteriorated the weld quality, where lack of penetration, underfill and weld centreline cracking were observed. A fast welding speed of resulted incomplete penetration and root porosity. Changing the focal position had the effect of changing the incident laser power density. Underfill and undercut defects was reduced by welding with filler metal.

Welding with filler metal reduced the crack sensitivity of AA 2024-T3 but it was also important to optimise the filler metal feed rate to avoid the formation of welding defects and keyhole instability. High feed rates produced instabilities, whereas, low feed rates did not sufficiently modify the chemical composition of the weld pool. The optimum feed rate was found to be around 2-3 m/min which minimised the crack sensitive range of Al-Mg, Al-Mg2Si and Al-Si according to the measurements obtained by EDX.

112 Due to the short wavelength of fibre laser, only weakly ionised laser induced vapour plume was formed rather than strongly ionised plasma. Therefore, both argon and helium shielding gases were found to be effective when welding AA 2024-T3 using fibre laser. The weld quality was improved using helium because the plume effect was reduced due to the higher ionisation potential of helium which minimised porosity formation. However, at a high laser power, undercut and underfill were smaller when using argon due to the lower energy transferred to the weld compared to helium.

3.5.2 Ti-6Al-4V Rapid heating and cooling of fibre laser welding induced a fine martensitic microstructure in the FZ and mostly martensitic and acicular α structure in the adjacent HAZ. The HAZ next to the BM consisted of primary α, intergranular β and martensite. The BM consisted of primary α and intergranular β.

Heat input supplied to the workpiece was mainly controlled by laser power and welding speed, which in turn, influenced the weld microstructure by modifying the peak temperatures experienced and, the heating and cooling rates. Decreasing the laser power or increasing the welding speed resulted in finer martensite and prior β grains in the FZ, whereas, increasing the laser power or decreasing the welding speed resulted in acicular martensitic structure and larger prior β grain size, as well as the formation of diffusional transformation constituents such as acicular α and grain boundary α. No significant differences in weld microstructure were observed over the range of focal positions investigated.

The weld top and bottom width all increased with increasing laser power, focal position and decreasing welding speed. Incomplete penetration or narrow root width were the main problems at low laser powers and fast welding speeds, whereas, undercut was the main defect at high laser powers. In general, the weld quality produced was good for all welding speeds as long as it was not too fast. Spatter was observed at the bottom surface at very fast welding speeds. Reinforcement increased with increasing laser power or welding speed but was not critical enough to affect the weld quality. Porosity detected was below the critical size. The root width increased more at negative focal positions than at the equivalent positive focal positions due to an enhanced interaction between laser and the material being welded with negative defocusing. It was therefore, possible to weld Ti-6Al-4V at high laser powers and welding speeds, over a wide range of focal positions, with characteristics and qualities that are pertinent to aerospace applications

113 4 EFFECT OF WELDING ON MECHANICAL PROPERTIES OF AA 2024-T3 AND TI-6AL-4V

4.1 Introduction The excellent beam quality, high power and shorter wavelength are few of the characteristics of fibre laser which are beneficial in the production of high quality welded joints of aluminium and titanium alloys. However, the welding parameters have a significant influence on the weld mechanical properties. This is especially true for the heat treatable AA 2024-T3 and Ti-6Al-4V alloys because varying the input welding parameters such as laser power, welding speed, focal position and etc. result in different output weld quality, shapes and microstructures leading to considerable variations in local mechanical properties [31]. The weld quality not only affects its mechanical properties but also influences its behaviour under loading conditions. For example, its tensile and fracture behaviour are affected by numerous factors, in particular, the shape of the weld cross-section, the metallurgy of the fusion zone and the heat affected zone and on the presence or degree of welding defects [247]. These factors especially weld shape, material inhomogeneity and critical sized geometrical welding defects such as cracks, porosities and undercuts as discussed in Chapter 3, can heavily influence the mechanical properties of the welded component in service by acting as sources of stress concentration and crack initiation during load and for subsequent failure. It was previously found by Weeter [289] that for Nd:YAG and CO2 laser welded heat treatable 2xxx and 6xxx series aluminium alloys, the tensile strengths often only reached up to 60-80% relative to that of the base metal and also reduced elongation to failure of 1-3%. Consequently, the aim of this investigation was to optimise weld mechanical properties. In order to determine the combinations of welding parameters which lead to excellent mechanical properties, different methods and approaches were used to test laboratory scale specimens under different material conditions and loading situations, before going into large and expensive structural tests after welding a component. The knowledge was then transferred to larger samples and even components.

The majority of published work to date on weld mechanical properties were derived from tensile testing transverse weld specimens denoted weldment (i.e. samples consisting of BM, FZ and HAZ). However, according to Cao et al. [12] due to inconsistency in sample preparation and experimental methods among different researchers, which were not always provided, large variations in tensile properties were reported because different sample sizes were used. Characterising the weld mechanical properties is a complex process due to heterogeneous microstructure and composition of the weld comprising the FZ, the HAZ and the BM. While it is convenient to simply test at a macroscopic level on the overall welded joint for bulk

114 mechanical properties, this was not the case in this investigation because it was necessary to determine the properties of the different regions within the weld on a scale comparable to that of the microstructure. Therefore, to account for the heterogeneous microstructural regions and to measure the inherent property gradients induced by welding, different testing procedures were followed. Micro-hardness indentation test was used as it is a relatively simple testing method which provides useful estimates of local mechanical properties for the various metallurgical zones. Uniaxial tensile tests were also used at different length scales to further investigate the material properties at large strains up to failure. Standard welded tensile specimens were used to characterise the overall global mechanical properties of the welded joints. Micro-tensile specimens were used to resolve the spatial variation in weld microstructure at high magnifications and to derive local stress and strain curves [290]. As it was difficult to measure the deformation of miniature specimens directly using conventional strain gauge technique, digital image correlation (DIC) technique in combination with an optical microscope was found to be suitable for measuring micro scale full field surface deformations [291]. The local constitutive behaviours determined from micro-tensile testing were useful input parameters for numerical simulation on modelling the mechanical behaviour of welded structures [292].

4.1.1 Digital Image Correlation There are several techniques available to determine local material behaviour and mechanical properties variations in welds. These include digital image correlation (DIC), in-situ testing of miniature specimens extracted from different regions that constitute the weld, Gleeble weld thermo-mechanical simulation and instrumented micro-hardness indentation [293]. The advantages of DIC for measuring strain variation across welds compared to other methods include the ability to determine the constitutive behaviour anywhere within the displacement data field in both local and global size scales with strains ranging from 0.05 to over 100%, real time full field measurement without any contact, no uncertainty regarding whether the material tested is truly representative of the material in the weld and no requirement on prior knowledge of material properties [294].

2D DIC can used for quantitatively determining the heterogeneous strain field and characterising the deformation mechanism of welds [295]. Measurement of micro-scale localised deformation within different weld regions can also be made by combining the 2D DIC technique with high spatial resolution microscopes and a micro-tensile stage. The errors in the global average strain and local strain variations are limited to accuracy of strain measurement of 10-4 [296,297]. The accuracy of DIC results depends on several parameters such as speckle size and density, subset size, step size and correlation criterion. The subset size is the dimension of one single square groupings of pixels and the step size is the distance between

115 the centres of adjacent subsets. The spatial resolution can be optimised by reducing the step size smaller than the subset size to overlap adjacent subsets, assuming that they always remain next to each other during the deformation. The influence of speckle pattern on the measurement accuracy was studied by Lecompte et al. [298] who showed that for small subset sizes, better accuracies can be obtained with smaller speckle sizes and that the optimal speckle coverage lies between 40-70% of the total surface area. Also, according to Hung and Voloshin [299], speckle size less than two pixels would cause greater uncertainties in speckle location when compared to larger ones. However, when the size becomes too large, the ability to detect smaller strains deteriorates.

There is currently no published work on the application of the DIC in mechanical testing of fibre laser beam welded AA 2024-T3 and Ti-6Al-4V. There are difficulties in applying the findings from previous works because these investigations were conducted under specific combinations of materials, welding processes and experimental procedures relevant to their research and in many cases, the welding process was not optimised so welding defects were frequently observed. Therefore, it was necessary to determine the influence of fibre laser beam welding on mechanical properties of AA 2024-T3 and Ti-6Al-4V welds to optimise the weld quality, by using the DIC for tensile testing as well as other conventional techniques such as micro-hardness indentation and scanning electron microscopy (SEM).

4.2 Materials and Experimental Procedures The same materials, AA 2024-T3 and Ti-6Al-4V, and welding process as illustrated in Section 3.1 were used for mechanical analysis. The welded specimens used for mechanical testing were butt welds instead of bead on plate welds. The butt joints were produced by placing two sheets 200 mm wide and 250 mm long each with a thickness of 3 mm for AA 2024-T3 and 2 mm for Ti-6Al-4V in the same plane, butted against one another with no gap. For mechanical testing of T-joints, fillet welds were produced using the welding parameters described in Section 3.2.5 for AA 2024-T3 and Section 3.3.4 for Ti-6Al-4V, by placing two parts at right angles to each other, forming a T-shape and welding the corners. The workpieces were clamped close to the weld line to avoid problems with misalignment and distortions, without compromising the joint access. Welded specimens for micro-hardness testing were prepared using the same method in Section 3.1.3 for metallographic specimen preparation.

Examination and testing of welds were conducted in accordance with acceptance level C and D of BS EN ISO 15614-11:2002 standard for electron and laser beam welding [239], which defines the conditions for the execution of welding procedure qualification tests. While it is compulsory to conduct radiographic, ultrasonic examination and surface crack detection in acceptance level C, it is not in level D so only visual and metallographic examinations were

116 conducted as in Section 3. It would be necessary to also perform these tests on the final welded component. In contrast, there is no specification on mechanical testing of welds in acceptance level D so the criteria in level C were followed. Therefore, micro-hardness testing on weld cross-sections and tensile testing of welded specimens were conducted.

Table 8 Examination and tests for welds in accordance with acceptance level C and D

Extent of examination and test Type of examination and test Level C Level D Visual examination 100 % 100 % Radiographic examination 100 % if required Ultrasonic examination 100 % if required Surface crack detection 100 % if required Butt weld Metallographic examination 1 section minimum 1 section minimum Hardness test if required N/A Transverse bend test if required N/A Longitudinal bend test if required N/A Transverse tensile test 2 specimens N/A Visual examination 100 % 100 % Ultrasonic examination 100 % if required Surface crack detection 100 % if required T-joint Metallographic examination 1 section minimum 1 section minimum Hardness test if required N/A Other tests if required N/A

Vickers micro-indentation hardness measurements were performed to determine the effect of fibre laser welding on the extent of the HAZ and hardening or softening of the FZ due to microstructural transformation. The variation of hardness over different microstructural constituents of weld cross-sections were measured. The Vickers hardness test uses a diamond-pyramid indenter with an angle of 136° between opposite faces of the diamond. Vickers hardness number (HV) is defined as the applied load divided by the surface area of the indentation calculated from average length of the diagonals of the indent. Experiments were conducted in accordance with ASTM E384 for micro-indentation hardness tests [300] and BS EN ISO 22826 for hardness testing of narrow joints welded by laser and electron beam [301]. The Vickers micro-hardness measurements were performed on weld cross-sections perpendicular to the weld line using a Zwick Roell Z2.5 (ZHU 0.2) hardness testing machine at a load of 100 g and a dwell period of 15 seconds at a speed of 60 μms-1 to characterise the whole hardness profile across the weld seams up to the base metal. The measurements were taken along three lines for AA 2024-T3 and two lines for Ti-6Al-4V. The number and spacing of indentations were chosen to maximise the number of hardness measurements in the narrow FZ and HAZ to accurately define the hardened or softened regions due to welding, while ensuring high accuracy in the hardness values. To avoid work hardening contributions from the adjacent indents, indentations were separated by 200 μm and the lines of indentation were separated by 500 μm. The minimum recommended spacing between the centre of any

117 indentation and the edge of the specimen for nonferrous metals must be at least three times the mean diagonal length of indentation for the Vickers hardness test according to ASTM E384 and BS EN ISO 22826 [300,301].

As it was not possible to extract standard round tensile specimens from specific regions within the overall microstructure of narrow laser welded joints, standard and micro flat cross-weld tensile specimens were used instead to determine the intrinsic local mechanical properties [29]. The detailed design of tensile test specimen used can be found in Appendix B. Two tensile specimens of 50 mm gauge length were prepared for each welding condition according to ASTM E8 [302], BS EN ISO 6892-1 [303] and AWS B4.0 [304] standards, with the weld line at the centre of the specimen perpendicular to the tensile axis. Uniaxial tensile tests were conducted at ambient temperature on a 200 kN Instron electromechanical universal testing machine in ram displacement control at a constant crosshead speed of 1 mm/min and an axial clip-on strain gauge extensometer Instron 2630-107 with 25 mm gauge length and maximum strain in the range between -10% and +100% was used to record strains.

Three different types of tensile tests were performed to evaluate the mechanical properties of fillet welded T-joints under different loading conditions. The first test applied tension in y direction with the weld line perpendicular to the tensile axis to simulate the hoop stress which corresponds to circumferential load in a pressurised fuselage skin. The second test applied tension in y direction with the weld line parallel to the tensile axis to simulate the tension stress in the upper fuselage [5]. The stiffener was not removed from the skin so a bending moment due to asymmetry of the stiffeners with respect to the loading plane was expected which would lead to the stiffeners being pulled off the skin.

The third test applied tension perpendicular to the plane of contact between the stiffener and the skin to verify the adhesion of the stiffener to the skin [148]. Such a load distribution is not expected in a real fuselage but the test was performed as a quality control of the weld quality rather than as a structural test [5]. In order to conduct the stiffener tension tests, an additional clamping system as shown in Figure 57 was used to hold the specimens. The specimens were machined from fillet welded T-joint plates with the dimensions given in Figure 57 and fixed in position by tightening to the clamping system via bolts using threaded pins which pass through the holes in the skin section. The stiffener section of the specimens was clamped by the top wedge grip and the clamping system was clamped by the bottom wedge grip of the tensile testing machine to tear off the stiffener from the skin.

118 Region of interest

Figure 57 An additional clamping system adopted to perform the test on the adhesion of the stiffener to the skin

Micro-tensile tests were also performed using a Deben in-situ dual lead screw micro-tensile testing stage with maximum loading to 2 kN at a constant displacement rate of 0.2 mm/min using a stepping motor. The tensile module was fitted to an optical microscope and the deformation during tensile testing was captured as digital images composed of 1600 x 1200 pixels every 0.5 seconds using the built in Zeiss Axiocam microscope CCD camera at 50x magnification. An initial load of 1 N was applied to avoid the influence of any initial nonlinearity in the output of the miniature test. Miniature flat welded tensile specimens with a gauge width of 5.0 mm, gauge thickness of 0.4 mm and gauge length of 8.0 mm, were wire cut from welded joints using electrical discharge machining (EDM) and then ground, polished and etch using the same methods illustrated in Section 3.1.3.

Scanning electron microscope (SEM) was used to examine the fracture surface of selected tensile specimens after tensile testing in the base metal and the welded joints at various magnifications to determine the fracture behaviour and the presence of welding defects.

All tensile tests were conducted in combination with a local strain field analysis using the ARAMIS 2D DIC 5M optical deformation analysis system developed GOM GmbH in order to characterise the weld mechanical properties. The DIC technique can be used to measure strains at various length scales so it was found to be ideal for testing miniature specimens as direct measurement of the deformation using conventional clip gauge technique was difficult [291]. In the applications of image analysis using the DIC, random patterns on the specimen surfaces were required [305] so the specimens were prepared by applying matt white background to the surface and then over spraying with a random black speckle pattern to improve contrast and was then used to define the subset for the processing steps to enable data acquisition by the DIC. The speckle diameter used in this study was around 0.3 mm which was calculated to be on average 10 pixels per speckle. However, for testing at high magnifications, the polished and etched weld surface topography which contained

119 characteristic features such as grain structures, and an adequate grey level contrast with the background was used directly as the random patterns for the correlation calculations [306].

During tensile tests, a series of images of the observed area were captured at different stages from the beginning to the end for strain field calculations by correlating the captured images at different stages. Since the weld consisted of different microstructural regions, it was expected that they exhibit different mechanical properties. The local strain evolution in each region was calculated by identifying their initial location on the reference image of the specimen and recording the strain field for each image during deformation. Local stress data were obtained from the globally applied load, assuming an iso-stress condition and that stress triaxiality does not influence the mechanical behaviour. Local strain data were obtained from the ARAMIS software for any position within the displacement field. The global stress data were combined with the local strain data to determine the local mechanical properties of the different regions.

For accurate displacement estimation, it had to be ensured that the front surface of the specimen was flat and in the same plane parallel to the camera and that out of plane motion was small enough to be neglected. In addition, to provide suitable spatial resolution and reliable results, the subset and step sizes were optimised for each specimen. The subset size had to be sufficiently large to resolve the distinctive strain distribution for the area of interest, facilitating a reliable correlation analysis with less random noise, but also small enough to distinguish strain differences in certain small regions and avoid averaging effects. The step size was changed to control the density of the analysed data where a larger step size generates coarser but quicker results [307].

4.3 Results and Discussion on AA 2024-T3

4.3.1 Micro-indentation Hardness The influence of fibre laser welding on microstructural transformations and variations of local hardness profiles of AA 2024-T3 welds was evaluated by measuring hardness in the direction perpendicular to the transverse weld cross-sections along three lines located 0.5 mm away from the top and the bottom surfaces, with a 1 mm gap between them. Micro-hardness testing in the transverse weld bead cross-sections as illustrated in Figure 58, Figure 59 and Figure 60, showed that all AA 2024-T3 welds were under-matched with the lowest hardness in the FZ. The hardness in the BM was the highest as expected. It was also found that the hardness in the HAZ was greater than in the FZ but lower than in the BM. Micro-hardness increased as a function of the distance from the weld centre in which the FZ had a hardness of around 90- 100 HV, the HAZ hardness in the range of 100-120 HV and the BM hardness in the range of 130-140 HV. The HAZ adjacent to the FZ showed hardness values close to that in the FZ

120 whereas, the HAZ adjacent to the BM showed a hardness close to that in the BM. Since the extent of the FZ and the HAZ was very small, the resulting hardness gradient was very steep. On the other hand, the hardness distribution was relatively uniform across the FZ in most specimens.

According to Cao et al. [12] the loss of strengthening precipitates or alloying elements and softening in the FZ, and over-aging in the HAZ were the main causes of hardness degradation in heat treatable aluminium alloys during welding process. The effect of grain growth with respect to strength was of minor importance but instead was mainly influenced by modification of precipitates [244]. Softening in the FZ was caused by microstructural changes as a result of very high temperatures experienced in the FZ and the associated rapid heating and cooling rates. The heating action of the laser led to segregation of elements, formation and growth of non-strengthening coarse precipitates, dissolution of strengthening precipitates and uniform distribution of precipitating elements during heating which then froze due to fast cooling rates. The hardening effect was therefore removed and the mechanical properties of the weld degraded. The hardness in the FZ was similar to the hardness measured in a fully solution treated and quenched AA 2024 of around 80 HV [308]. Even though the FZ partially recovered its hardness by natural ageing at room temperature for several days after welding, the effect was small due to inhomogeneous distribution of solute atoms. Loss of volatile elements such as magnesium and zinc for strengthening also contributed to lowering the hardness in the FZ by affecting the weld pool chemistry [11] as previously discussed in Section 3.2.3.

The welding thermal cycle also affected the precipitation behaviour in the HAZ such as dissolution, precipitation and coarsening so the HAZ was divided into two different microstructural regions of partially melted zone and over-aged zone. The hardness in the partially melted zone decreased due to dissolution of strengthening precipitates during melting and segregation of alloys during solidification. In the over-aged zone, coarsening of the strengthening semi-coherent S´ phase as well as transformation to the non-strengthening incoherent stable S phase reduced the hardness [21].

The amount of softening in the FZ and the HAZ depended on heat input. Higher heat input was achieved by either increasing laser power or decreasing welding speed. Since changing focal position and filler metal feed rate only had a small influence on the heat input as shown in Figure 59 and Figure 60, the effect of these two parameters on micro-hardness was relatively small but mainly affected the weld width, where increasing the feed rate or focal position either positive or negative also increased both the face and the root weld widths. The detailed measurement results can be found in Appendix B.

121 200 W93 ~ W96 P= 1.9 kW, f = +4 mm, Ar 175 V = 1.0 ~ 1.7 m/min

150

125

100 Vicker'shardness (HV0.1) 1.0 m/min 75 1.3 m/min 1.5 m/min 1.7 m/min 50 -3 -2 -1 0 1 2 3 Distance from centre (mm)

Figure 58 Effect of welding speed on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds

200 W17 P= 4.9 kW, V = 5.0 m/min, 175 f = +4 mm

150

125

100 Vicker's hardness hardness Vicker's(HV0.1) Top 75 Mid Bot 50 -3 -2 -1 0 1 2 3 Distance from weld centre (mm)

Figure 59 Effect of wire feed rate on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds

200 W72 P= 3.9 kW, V = 2.0 m/min, 175 f = -4 mm, Ar

150

125

100 Vicker'shardness (HV0.1) Top 75 Mid Bot 50 -3 -2 -1 0 1 2 3 Distance from centre (mm)

Figure 60 Effect of focal distance on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds

122 In fact, it was found that as the heat input increased the micro-hardness decreased. Decreasing the welding speed led to a wider weld width and a lower hardness of approximately 90 HV as shown in Figure 58 b) and c) compared to 100 HV at high welding speeds as shown in Figure 58 g) and h), due to lower cooling rate and longer time exposed to high temperatures with increasing heat input.

The reduction in hardness at lower welding speed was also attributed to the increased the dendrite cell and grain sizes and the coarse secondary phase in the FZ and the HAZ due to the relatively slower cooling rate [136]. On the other hand, for the specimens which had smaller heat input and consequently, a faster cooling rate and shorter duration at high temperatures arising from a smaller weld pool volume, the hardness was greater in the FZ and the HAZ because of finer grain size [154]. For the same reason, the loss of low boiling point elements for strengthening such as magnesium was minimised by the shorter interaction time and lower laser power density [142]. However, the use of very low heat input was avoided as it would induce a crack susceptible microstructure with poor toughness.

Post weld artificial ageing heat treatment only without prior solution treatment has insignificant effect on the hardness recovery in the FZ and the HAZ [244,309]. The over-aged zone in the HAZ which had the maximum amount of incoherent non strengthen S phase did not respond well to post weld ageing [242]. In order to considerably or fully recover the hardness in these regions to the same level as the BM, it was necessary apply a solution treatment followed by quenching before applying subsequent natural (T3) or artificial ageing treatment. This was also the key reason why preheat was not used when welding heat treatable aluminium alloys due to the possibility of over-ageing but rather, the time at high temperatures was minimised to maximise the mechanical properties of the AA 2024-T3 welds.

4.3.2 Global and Local Tensile Properties Macro- and micro-tensile specimens were obtained from butt-welded 3 mm thick AA 2024-T3 sheets produced under numerous welding conditions of laser powers, welding speeds, focal positions and welding with or without filler metal. Digital image correlation (DIC) technique was used to capture digital images of the deforming surface of the specimens during uniaxial tensile deformation every second to compute the corresponding displacement and strain. Table 9 lists the welding parameters of some of the standard tensile specimens which were tested using the DIC, with the laser power in the range between 2.9 and 4.9 kW, the welding speed between 1.5 and 3.0 m/min, the focal position of either 0 or +4 mm and autogenous welding or welding with different filler wire feed rates between 2.0 and 5.2 m/min.

123 Table 9 Differernt sets of welding parameters used for tensile testing macro-tensile welded specimens

Experiment Laser power Welding speed Focal position Wire feed rate number (kW) (m/min) (mm) (m/min) W85 4.9 3.0 0 - W86 3.9 2.0 0 - W87 3.9 2.0 +4 - W88 4.9 3.0 +4 5.2 W98 2.9 1.5 +4 2.0 W99 2.9 1.5 +4 - W100 2.9 1.5 +4 2.6 W101 2.9 3.0 +4 5.0

The joint efficiency was calculated for each specimen in terms of the parent material ultimate tensile strength of 420 MPa as shown in Figure 61. The joint efficiency represents a generic level of confidence in the overall strength of a weld seam relative to the base metal. A value greater than 100% indicates that the weld metal is stronger than (over-match) the base metal, whereas, less than 100% is weaker (under-match). There is no authoritative document citing specific data on which values for joint efficiency are based. However, the term was originally developed considering the results of burst tests of pipe performed by major pipe manufacturers and applied in the ASME B31.1:1935 Code for Pressure Piping in setting the allowed working stress of various seams [310]. In most cases, the optimum condition required for a full penetration weld is to match the strength of the weld to that of the BM, whereas, over- matching is not often required. Under-matching the weld strength on the other hand, can cause strain concentrations and increase constraint in the FZ, which lead to confined plasticity development within the FZ and reduce the overall plastic straining capacity of the welded joint subjected to tension load by restricting crack growth to the weaker strength FZ and HAZ of the weld [11]. In addition, welding defects such as partial penetration, undercut, underfill and porosity can significantly influence the local stress field around the weld and lead to severe stress concentrations which reduce the strength [311]. The weld mechanical properties were therefore, a function of the microstructural characteristics developed in different regions of the weld and the sources for the lower strength of the FZ and the HAZ was discussed in Section 4.3.1.

It was found that fracture initiated at the softened FZ or HAZ of the weld where the strength was lower compared to that of the BM so the obtained tensile strength and calculated joint efficiency values were in terms of the tensile strength of the weaker region of the weld. In many cases, the joint efficiency of greater than 2/3 or 66% of the BM was determined to be acceptable for different aluminium alloys according to the work by several other researchers [26,312–314]. It was found that the joint efficiencies of the tested specimens as shown in Figure 61 were mostly well above 66% except for one specimen which was welded using a

124 laser power of 2.9 kW, welding speed of 3.0 m/min, focal position of +4 mm and a wire feed rate of 5.0 m/min. High weld strengths and joint efficiencies overall were the result of the fine microstructures developed in the FZ due to the low heat input and fast solidification found during fibre laser welding. Since high joint efficiencies were achieved in these specimens, a more stringent threshold of 85% was also used to assess their strength. The lowest joint efficiency values were obtained in the three specimens welded using the fastest welding speed of 3.0 m/min. The values greater than 85% were achieved in the specimens welded using the lowest laser power of 2.9 kW and also the slowest welding speed of 1.5 m/min and the maximum value was obtained by welding with filler wire at a feed rate of 2.0 m/min. In summary, the joint efficiency was good as long as the heat input was not too low or the welding speed was not too fast

100 Weld acceptable when UTS above 85% of PM UTS (420 MPa) 90

40 Joint efficiency Joint efficiency (%) 30

0 W85 W86 W87 W88 W98 W99 W100 W101

Figure 61 Joint efficiencies of welded macro-tensile specimens for the corresponding welding parameters listed in Table 9

(i) In order to evaluate quantitatively the differences in mechanical properties among different welds, the development of strain distribution during loading was determined by the DIC. Figure 62 shows the full field y strain distribution for four different welding conditions obtained from standard tensile testing at different percentage of the fracture time at 25, 50, 75 and 99%. The colour scale bar which represented the level of strain in the maps was independently chosen for each set of welding parameters to maximise contrast and different colours indicate different strains prior to failure. The strain profiles across the weld showed a sharp strain gradient from the weld centreline to the BM at the onset of final fracture. The maximum strain localisation was located in the HAZ for the top two specimens welded using a laser power of 3.9 kW and a welding speed of 2.0 m/min (W86, W87), and in the FZ for the other two bottom specimens welded using a laser power of 2.9 kW and a welding speed of 1.5 m/min (W98,W99), while the minimum occurred in the BM. Deformation in the specimens which failed in the HAZ was uniform until 75% of the fracture time, after which localisation occurred on one side of the HAZ

125 up to failure with local strains of around only 3% for the specimen with a focal position of 0 mm and 8% for the specimen with a +4 mm positive defocusing. On the other hand, the strain distribution in the specimens which failed in the FZ was uniform and symmetrical about the weld centreline throughout the deformation with relatively larger local strains of around 11% in the FZ for both specimens welded with or without filler metal, compared to the other two specimens.

The strain measured in the BM was of the order of only 0.5-1.0% at failure which indicated that it was still under elastic loading. Although the boundaries of different characteristic regions corresponding to the BM, the HAZ and the FZ cannot be easily identified from the strain maps, their locations were measured and marked outside the processed regions prior to testing and also it was possible to determine their extent by examining the highly non-uniform strain distribution across the weld at different load levels [315]. It was obvious from these strain maps that the stiffness of the welded specimen was a result of the stiffness of the three different microstructural regions

126 The global tensile behaviour of the welded specimens and the BM and the corresponding mechanical properties are shown in Figure 63. It was possible to create any size and number of gauge lengths on the processed DIC images and so the global stress and strain curves were determined for a 25 mm gauge length equal to that of the extensometer used. For mechanical characterisation of the welded joints, the elastic modulus, yield strength, ultimate tensile strength and elongation to failure were determined. Global tensile test results showed significant losses in ductility and tensile strength in the welded specimens compared to the unwelded BM due to plastic strain localisation and increased constraint within the lower strength weld region of the welded joint for the composite gauge length. Only small variations in the yield strength and elastic modulus were observed, while considerable differences in elongation to failure and ultimate tensile strength were measured. These two properties were the lowest in the specimen welded using a laser power of 3.9 kW and a welding speed of 2.0 m/min and a focal position of 0 mm (W86), partly because relatively deep undercut was formed in this specimen as previously identified in Section 3.2.2.

When the focal position was 0 mm, the power density was the maximum and the heat was localised into a small spot, producing a narrow weld and also increased interaction time to promote grain growth in the FZ. Defocusing the laser beam reduced the power density and led to the formation of slightly finer grains in the FZ which resulted in superior weld tensile properties. Positive defocusing at +4 mm while keeping the remaining welding parameters identical (W87) improved the ductility and tensile strength. However, reducing the power density increased the solidification cracking sensitivity at the same time and therefore, small weld centreline solidification cracking was observed in this specimen. Further improvements were found in the specimens welded using a reduced laser power of 2.9 kW and welding speed of 1.5 m/min while maintaining the +4 mm defocus, where cracks were avoided and undercut was small enough to be insignificant. The addition of filler metal (W98) reduced the risk of welding defects and resulted in a significantly greater ductility over 3.5% and a fairly higher tensile strength of around 380 MPa than the rest of the specimens.

127 a) b) 600 600

500 500

400 400

300 300

200 True stress(MPa) 200 Engineering Engineering stress(MPa) BM BM 100 W86 (P=3.9 kW, V=2.0 m/min, f=0 mm) 100 W87 (P=3.9 kW, V=2.0 m/min, f=+4 mm) W86 (P=3.9 kW, V=2.0 m/min, f=0 mm) W98 (P=2.9 kW, V=1.5 m/min, f=+4 mm, filler) W87 (P=3.9 kW, V=2.0 m/min, f=+4 mm) W99 (P=2.9 kW, V=1.5 m/min, f=+4 mm) W98 (P=2.9 kW, V=1.5 m/min, f=+4 mm, filler) 0 0 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Engineering strain True strain c) d) 4.0 70 450 450 Elongation to failure Elastic Modulus Ultimate tensile strength Yield strength 400 400 3.5 60 350 350 3.0 50 300 300 2.5 40 250 250 2.0 30 200 200 1.5

Yield Yield strength (MPa) 150 150 Elastic Elastic modulus(GPa) Elongation Elongation to failure (%) 20 1.0 100 100 Ultimate tensile strength (MPa) 10 0.5 50 50

0.0 0 0 0 W86 W87 W98 W99 W86 W87 W98 W99

The influence of local material behaviours in the various weld zones on the overall weld response was determined using the DIC by assuming an iso-stress condition for all specimens, where the global stress was considered as the corresponding local stress at any point within the analysed displacement data field. The local strain data were derived from the DIC by measuring the local strain over each individual region, which were then plotted against the global stress data to obtain the local tensile properties in the FZ and the HAZ on either side of the FZ (HAZ1, HAZ2) as shown in Figure 64 [293]. As it was only possible to obtain the full tensile response in the weakest region where strain localisation occurred, the stress and strain curves in the stronger regions such as the BM were not obtained. The high hardening rate and low ductility observed in the global stress and strain curves of the welded specimens proved that strain localisation occurred in the weaker regions of the weld. In fact, as all the AA 2024- T3 specimens were under-matched, the local strain evolution in the FZ or the HAZ was successfully calculated up to complete fracture, which was not possible to determine from

128 standard tensile specimens without the DIC because of the narrow size of the FZ. It was found that the y strains measured in the weld was much higher than the measured global fracture strains which indicated that the strain distribution within the 25 mm gauge length was not uniform but highly localised in the weaker FZ or HAZ so the overall behaviour was dominated by that of the weakest component of the specimen and minimal plastic deformation occurred outside the weld [294]. The maximum strain was reached in one of the HAZs for the specimens W86 and W87, whereas, it was reached in the FZ for the specimens W98 and W99. As the property gradients were less steep in the specimens W87, W98 and W99 with wider weld widths than W86 due to +4 mm defocusing, greater plastic deformations were obtained in general as well as in the stronger microstructural regions before fracture. The local strain measurements obtained from the DIC were useful for providing the input weld material parameters which take into account the deterioration of the material properties of the FZ and the HAZ for numerical simulations on global models of welded structures.

P=3.9 kW, V=2.0 m/min, f=0 mm P=3.9 kW, V=2.0 m/min, f=+4 mm 450 450

400 400

350 350

300 300

250 FZ 250 FZ HAZ1 HAZ1 200 200 HAZ2 HAZ2

150 150 True stress (MPa) stress True True stress (MPa) stress True Global Global 100 100

50 50

0 0 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 True strain True strain

P=2.9 kW, V=1.5 m/min, f=+4 mm, filler P=2.9 kW, V=1.5 m/min, f=+4 mm 450 450

400 400

350 350

300 300

250 FZ 250 FZ HAZ1 HAZ1 200 200 HAZ2 HAZ2

150 150 True stress (MPa) stress True Global (MPa) stress True Global 100 100

50 50

0 0 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.00 0.02 0.04 0.06 0.08 0.10 0.12 True strain True strain

Figure 64 Local mechanical responses in FZ and HAZ constructed from full field DIC tensile tests compared to overal responses for four different welding conditions

129 (ii) Scanning electron microscopy (SEM) analysis was performed on the fracture surfaces of the AA 2024-T3 DIC tensile specimens as shown in Figure 65 to examine the fracture behaviour and the existence of welding defects such as hot cracks and porosities. It was observed that for all specimens inspected, failure occurred within the weld, either in the FZ or in the HAZ/FZ boundary but not in the BM due to weld under-match. Different characteristics of fracture morphology were observed among these specimens welded using different welding parameters. For the specimen which was welded using a laser power of 3.9 kW, a welding speed of 2.0 m/min and focal positions of 0 mm (W86), a fracture surface characterised by tearing ridges due to transgranular cleavage, typical of brittle fracture, as well as a few dimples which are the characteristic feature of ductile fracture, were observed. A large amount of hydrogen induced spherical shaped pores of different sizes similar to those observed by Alfieri et al. [158], ranging from around 100 to 300 μm with round tip dendrites, were identified on the fracture surface in clusters which were responsible for crack initiation during deformation. Porosities reduced the effective cross-sectional area of the welded joints and therefore, caused stress concentrations which deteriorated the strength of the joints in proportion to the reduction of the cross-sectional area.

For the specimen welded with a focal position of +4 mm while keeping other parameters constant (W87), a rough and irregular fracture surface, consisting of microvoids and dimples that act as microvoid nucleation sites, was observed. This indicated that the specimen failed in a more ductile manner where fracture initiated by microvoid coalescence and then by dimple rupture. Rough fracture surfaces with fine equiaxed dimples are the characteristic features of ductile transgranular fracture mode. Micro pores were also observed in this specimen but as it was more ductile, its strength and ductility reductions due to porosity were less significant than in the brittle specimen. For the other two specimens welded using a lower laser power of 2.9 kW, a lower welding speed of 1.5 m/min and a focal position of +4 mm, even greater amount of finer dimples was found to dominate the fracture surfaces which is the characteristic feature of purely ductile fracture. A significant amount of localised microscopic weld plasticity was observed in these specimens which improved the tensile strength and ductility of the respective welds as previously determined from tensile testing. While intergranular inter- dendritic micro hot cracks and porosities were detected in the three autogenous welds, with a slightly higher amount in the first two specimens, they were almost or completely absent in the specimen welded with a filler metal (W98) so a significant improvement in tensile strength and ductility were obtained in the specimen welded with a filler metal. In fact, porosity level less than 3% of the total volume does not affect much the yield and tensile strength of a material but may affect ductility [316]. Although the specimen welded without a filler metal contained

130 some welding defects (W99), it still failed in a ductile manner similar to that in the specimen welded at a higher laser power and welding speed (W87).

Figure 65 SEM fracture morphology of the DIC tensile test specimens for the four different welding conditions at 500x magnification showing different modes of failure and presence of welding defects

131 (iii) Micro-tensile testing was conducted using an optical microscope at 50x magnification and the etched metallographic surfaces as the random patterns of adequate contrast for the DIC, to determine the influence of welding parameters on the mechanical properties of the welded joints. The small and varied size of different microstructural regions of the weld made it difficult to accurately determine the local mechanical properties from standard tensile specimens. Also, since the plastic strain distribution in the standard tensile specimen was inhomogeneous over the gauge length, the tensile properties were dependent on the chosen gauge length at which they were measured [29,317]. The advantage of using micro-tensile specimens was that it was possible to obtain a close up view of the entire weld zone at high magnifications and resolution and also observe the deformation directly from the clearly visible etched FZ and HAZ without the need to apply speckle patterns on their surfaces. The difficulties in accurately measuring strain in the micro-tensile specimens due to their limited size was solved by using the DIC which allowed to determine both global and local strain field of each individual regions from the examined surface area. It was therefore necessary to evaluate the reliability of the results of the micro-tensile specimens by comparing with those of the standard tensile specimens as shown in Figure 66.

The variations in tensile responses of the micro-tensile specimens were consistent with the results of the equivalent standard specimens W86, W87, W98 and W99 as shown in Figure 63 so they proved to be suitable for determine the mechanical properties of the welded joints. However, there was still some small differences in strength and elongation values between the micro- and standard tensile specimens, where the micro-tensile specimens exhibited around 10% lower values most likely due to constraint effect from reduced thickness similar to nearly plane stress conditions experienced in the thinner 0.4 mm thick micro-tensile specimens than in the full 3 mm thick standard tensile specimens, and also from stronger adjacent regions [318].

Figure 66 shows first the effect of changing laser power on the global mechanical properties of the welded joints, in the range between 1.5 and 2.6 kW for the specimens welded using a constant welding speed of 2.1 m/min, a focal position of 0 mm and argon shielding gas. It was found that increasing the laser power also increased the heat input into the workpiece so a coarser microstructure was produced in the FZ and the weld width became wider. A wider weld pool size reduced the risk of solidification cracking and decreased the stress levels across the weld seam during solidification. As a result, a trend was found where yield strength decreased but tensile strength and elongation to failure increased with increasing laser power [43,44,137]. Secondly, the effect of changing welding speed from 2.7 to 5.1 m/min was examined from the specimens welded using a constant laser power of 4.5 kW, a focal position of 0 mm and argon shielding gas. Increasing the welding speed led to a decrease in heat input

132 and resulted in the formation of finer dendritic structure and grain sizes in the FZ, minimise microstructural modification or loss of strength by dissolution or coarsening of the strengthening precipitates in the HAZ and reduced low boiling point alloying element evaporation due to faster cooling rate and shorter duration at peak temperatures for grain coarsening [319]. The columnar grains in the FZ grew normal to the welding direction at higher welding speeds, whereas, they curved away from the normal to the welding direction and aligned themselves with the welding direction at lower welding speeds [43,44]. In addition, increasing the welding speed led to narrower weld width and increased the risk of trapping porosity at relatively fast welding speed of around 5.0 m/min as discussed in Section 3.1.1 which was not fast enough to prevent nucleation of pores but also not slow enough to allow porosity to escape. The grain growth was limited at high welding speeds so the yield strength increased, whereas, the elongation to failure and the tensile strength decreased with increasing welding speed [31,43,44,137]. a) b) 400 450 V=2.1 m/min f=0 mm, Ar 350 400

350 300 300 250 250 200 200 W86 (Micro)

150 1.5 kW True stressTrue (MPa) Truestress (MPa) W87 (Micro) 150 W98 (Micro) 1.7 kW 100 W99 (Micro) 100 W86 (Macro) 1.9 kW W87 (Macro) 50 2.3 kW W98 (Macro) 50 W99 (Macro) 2.6 kW 0 0 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 True strain True strain c) d) 450 450 P=4.5 kW, f=0 mm, Ar P=3.0 kW, V=2.0 m/min, Ar 400 400

350 350

300 300

250 250

200 200

0 mm True stressTrue(MPa) 150 Truestress (MPa) 150 2 mm 2.7 m/min 100 3.3 m/min 100 4 mm 3.9 m/min -2 mm 50 4.5 m/min 50 5.1 m/min -4 mm 0 0 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 True strain True strain

Figure 66 DIC computed global tensile response of micro-tensile specimens a) compared to macro-tensile specimens and for different b) laser powers, c) welding speeds and d) focal positions

133 Finally, the effect of changing focal position from +4 mm to -4 mm was studied in the specimens welded using a constant laser power of 3.0 kW, a welding speed of 2.0 m/min and argon shielding gas. Defocusing the beam reduced the power density and increased the weld width so the tensile strength and ductility increased while, the yield strength decreased with increasing focal position, in both positive and negative directions. While the overall heat input was unchanged by defocusing, since the area affected by the laser beam was larger, the wider weld width benefited its strength. Wider welds were found to be less crack sensitive than narrower welds since the occurrence of hot cracking was influenced by the rapid solidification rate which caused development of high thermal shrinkage strains and stress gradients that led to a high crack initiation rate. Therefore, a lower welding speed, a larger focal spot size and a larger laser power produced wider welds to lower the crack sensitivity by controlling the solidification [12].

Figure 67 shows the DIC full field strain maps just before rupture, local constitutive behaviour of the FZ and the HAZ as well as the global response, and the strain distribution across the weld seam at different percentages of the fracture time determined from the micro-tensile specimens at a welding speed of 2.7 m/min. The results for different welding speeds corresponding to those in Figure 66 can be found in Appendix B. As it can be seen, the strain fields along the gauge length were determined by analysing the etched surface of the specimens during testing. A gap in the strain maps were produced at high levels of plastic deformation, where cracks started to form and grow in the FZ, significantly affecting the subset geometry and making it no longer possible to identify the subsets in the subsequent image series to calculate the strain field. It was found that in all tested specimens, the cracks initiated from the FZ. At lower welding speeds, the specimens fractured in the FZ, while at higher welding speeds, the cracks propagated towards the fusion boundary where the property gradients were the steepest and a mixed mode failure in both the FZ and the HAZ was observed. As discussed above, the tensile strength and the ductility decreased with increasing welding speed so the maximum local plastic strain distribution within the weld was also found to decrease with increasing welding speed from around 8% at 2.7 m/min to around 5% at 5.1 m/min.

134

350 8 t=25% 7 300 t=50% t=75% 6 250 t=90% 5 t=95% 200 t=99% 4

150 Strain (%)

3 True stressTrue (MPa) 100 2 FZ HAZ 50 1 Global 0 0 0.00 0.02 0.04 0.06 0.08 0.10 -3 -2 -1 0 1 2 3 True strain Distance from weld centre (mm)

(iv) Stiffeners in T-joints were welded to the base plate to reinforce the stiffness and strength of welded structures. The T-joint specimens were tested under three different loading conditions to determine the difference in strength and ductility compared to butt welded specimens. The stiffener was not removed from the base plate prior to testing in order to examine its influence on the joint mechanical properties and to realise as close as possible the actual loading conditions in a real fuselage, where the stiffener induce a bending moment due to asymmetry with respect to the loading plane. Figure 68 shows the strain distribution along the gauge length of the longitudinal T-joint tensile specimens during loading to final fracture viewed from the front and from the side. The stiffener and the weld seams in the longitudinal specimens were parallel to the loading direction which represented the tension stress in the upper fuselage of the aircraft [5].

It was found that initially strain localised at the top of the specimen where the stiffener end started to separate from the base plate at the weld toe and progressed towards the centre of the specimen with a further increase in the applied displacement. The stiffener remained

135 attached to the base plate on the bottom half of the specimen where the strain in the loading direction remained below approximately 5%. Once the stiffener became separated all the way down to around the centre of the specimen, the specimen behaved like the BM and failed in the middle after a slight necking in the base plate at 45° relative to the tensile axis. Throughout the test, the strain levels close to 0% were measured on the stiffener side because the interface between the base plate and the stiffener became separated first and therefore, the load applied to the stiffener was minimal.

Front

50% 75% 99% 100%

Side

50% 75% 99% 100%

Figure 68 Processed and raw DIC image data of fibre laser welded AA 2024-T3 longitudinal T-joint tensile specimens at 50%, 75%, 99% and 100% of the time to failure

136 Transverse T-joint specimens were tested to simulate the load distribution of hoop stress in a pressurised fuselage skin perpendicular to the welding direction [5]. The stiffener and the weld seams in the transverse specimens were perpendicular to the loading direction. Figure 69 shows the strain distribution along the gauge length of the transverse T-joint tensile specimens during loading to final fracture viewed from the front and the side. The strain distribution around the weld seams was relatively small at 50% of the fracture time with only around 2% strain observed from both sides. At 75% of the fracture time, strain localisation at the weld toe was detected at higher strains of around 5% compared to 3% in the BM. Just before failure at 99% of the fracture time, it was clear that the maximum strain of around 7% occurred at the weld toe when viewed from the front and also higher strains developed within the BM adjacent to the top weld toe at 45° relative to the tensile axis. The lower strength of the weld seam caused a notch effect so the failure originated at the weld toe near the fusion boundary on the skin side and the crack propagated along the fusion boundary and finally into the skin through the entire thickness with an orientation of 45° from the tensile axis. Again, the strain distribution on the stiffener side remained close to 0% during the test, similar to that observed in the longitudinal specimens.

Figure 69 Processed and raw DIC image data of fibre laser welded AA 2024-T3 transverse T-joint tensile specimens at 50%, 75%, 99% and 100% of the time to failure

137

Figure 70 Processed and raw DIC image data of fibre laser welded AA2024-T3 T-joint stiffener tension specimen at 50%, 75%, 99% and 100% of the time to failure

Stiffener pull off testing applied tension into the stiffener of the T-joint specimens to evaluate the weld seam quality and adhesion of the stiffener to the base plate as a quality control [5]. Figure 70 shows that strain localisation occurred at the weld toe and as the applied displacement increased, a crack was observed starting at the weld toe which then propagated along the fusion boundary. The fracture took place when the crack propagated along the fusion boundary towards the other end of the stiffener which resulted in a complete separation of the stiffener from the base plate, whereas, the base plate did not fail due to the prior failure along the fusion boundary [163]. a) b) 600 12 200

500 10 150 400 8

300 6 100

Force(kN) Stress(MPa)

Truestress (MPa) 200 4 PM 50 Butt 100 2 Ideal Fillet (Long) Undercut Fillet (Trans) Porosity 0 0 0 0.00 0.05 0.10 0.15 0.0 0.5 1.0 1.5 2.0 2.5 True strain Displacement (mm)

138 Figure 71 a) shows that the elongation of both the transverse and longitudinally loaded T-joint specimens was reduced compared to the conventionally tensioned flat BM tensile specimen. This was due to the presence of weld seams in the T-joint cross-section, which even in a small proportion deteriorated the plastic deformation capacity of the base plate. For the longitudinal T-joint specimen (green), the measured yield strength was greater than the BM (blue) by around 40 MPa whereas, the elongation to failure dropped by 0.04. It was determined that the constraint effect from the weaker FZ and high stiffness of the stiffener along the loading direction were responsible for the increased strength and decreased fracture strain of the longitudinal specimens. The stiffener had an effect of reducing the hardening and necking ability of the longitudinal specimens. For the transverse T-joint specimens (black), since the stiffener did not contribute much to the strength of the specimen, the yield strength was lower compared to the BM and because the deformation was localised in the welded joint, the elongation to failure was significantly reduced due to stress concentration. However, the mechanical properties of the transverse T-joint specimens were superior to the butt welded specimens (red) because the fraction of the specimen cross-section occupied by the weld seam was smaller and therefore, the contribution of the BM on the global joint mechanical properties was greater.

All stiffener tension (pull off) specimens as illustrated in Figure 71b) showed a linear increase in force with displacement up to a certain point where a sudden drop in the measured load was observed as a result of the load being sufficiently large to cause the stiffener and the base plate to start separating. For the specimens which had no visible weld defects (blue), considerable amount of additional load was required to completely pull off the stiffener. In contrast, in the case of the specimens which contained a small amount of welding defects such as undercut (red) and porosity (green), only a small extra load was enough to completely pull off the stiffener from the base plate.

4.4 Results and Discussion on Ti-6Al-4V

4.4.1 Micro-indentation Hardness The influence of fibre laser welding on local hardness profiles of Ti-6Al-4V welds was evaluated by measuring hardness in the direction perpendicular to the transverse weld cross- sections along two lines located 0.5 mm away from the top and the bottom surfaces, with a 1 mm gap between them. All specimens were welded without filler metal, using a constant +4 mm defocus above top surface of the workpiece and argon shielding gas. Different combinations of laser power and welding speeds were investigated. The detailed measurement results can be found in Appendix B. Micro-hardness testing indicated that as shown in Figure 72, the FZ and the HAZ were harder than the BM. The variation in hardness

139 observed across the weld was consistent with the observed changes in microstructure, with a steep hardness gradient from the FZ to the BM. The lowest hardness was measured in the BM and the FZ showed the highest hardness due to the formation of needle like martensite, α´ in the solidified FZ as previously discussed in Section 3. The hardness in the HAZ decreased with increasing distance from the weld centreline due to a drop in the fraction of α´ from 100% in the FZ to 0% in the BM. The hardness of the HAZ adjacent to the FZ was close to that of the FZ because it had a similar α´ and acicular α microstructure whereas, the HAZ close to the BM showed a similar microstructure to the BM [320]. The lowest hardness value in the BM on average of around 330 HV, an intermediate value of approximately 350 HV in the HAZ and the highest value in the FZ on average of around 380 HV, were all in close agreement with literature values on laser beam welded Ti-6Al-4V [185,190,191].

The hardness was determined by the solidification microstructure in different regions of the weld, as a function of heat input, peak temperature and cooling rate. At fast cooling rates in excess of 410°Cs-1 from above the β transus temperature (of around 980°C) and martensite start temperature (of around 650°C), led to complete diffusionless transformation from β to metastable needle like martensitic microstructure [282]. Cooling at a rate not fast enough for diffusionless transformation resulted in acicular alpha formation and even slower cooling rate resulted in precipitation of alpha along the prior β grain boundaries [67]. The temperature in the FZ exceeded the β transus temperature and the cooling rate was sufficient to undergo diffusionless transformation. The increased hardness in the FZ was associated with the formation of martensite as a result of the high cooling rates. The hardness in the HAZ adjacent to the FZ was similar to that in the FZ because this region experienced temperatures greater than the β transus temperature and so the microstructure mainly consisted of martensite and acicular alpha. On the other hand, the HAZ adjacent to the BM was lower because the peak temperature reached was less than the β transus temperature but above the martensite start temperature to form a limited fraction of martensite and primary α and intergranular β. The equiaxed microstructure of the mill annealed BM in the α + β phase field consisted of equiaxed α with intergranular β but no α´, which resulted in a further decrease in hardness to a minimum in the BM.

As shown in Figure 72, on average, the hardness in the FZ decreased with increasing heat input. Higher heat input was achieved by either increasing the laser power or decreasing the welding speed. Changing the focal position also changed the incident laser power density but not the heat input. For lower heat inputs, the time spent at higher temperatures decreased and the cooling rate increased over the solidification temperature range, which promoted the formation of finer martensite and prior β grains and as a consequence, increased the FZ hardness [321]. On the other hand, for higher heat inputs, the time spent at high temperatures

140 increased and the cooling rate decreased, so the martensitic phase changed from needle like to plate like morphology, the prior β grain size increased and in addition, diffusional transformation constituents such as acicular α and grain boundary α appeared [205,321]. Therefore, increasing the laser power or decreasing the welding speed reduced the hardness in the FZ. It was found that the weld width of specimens increased with increasing laser power at a constant welding speed, or with decreasing welding speed at a constant laser power. Similar to the discussion above, the weld width was larger at higher heat inputs and the hardness gradient from the FZ to the BM was relatively smaller than for the specimens with narrower weld width. For the specimens with narrower weld width, variations in the hardness gradient was considerable especially at welding speeds greater than 3.0 m/min which caused larger differences in microstructures, with the maximum hardness in the FZ close to 390 HV. In contrast, the maximum hardness in the FZ measured in the specimens welded at welding speeds below 3.0 m/min was around 370 HV which was not significantly greater than the hardness in the BM and therefore, supported the arguments above on the relationship between the amount of heat input and the hardness. In summary, the greatest hardness in the FZ was achieved at higher cooling rates where fully martensitic microstructure was formed and then the hardness progressively decreased towards the BM as the fraction of martensite dropped, finally disappearing at the HAZ/BM interface.

450 P= 3.5 kW, V = 7.2 m/min f= +4 mm, Ar

400

350

Vicker'shardness (HV0.1) 300

Top Bot 250 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Figure 72 Micro-indentation hardness distributions of fibre laser welded Ti-6Al-4V autogenous welds

4.4.2 Global and Local Tensile Properties (i) Several macro-tensile specimens of the same welding conditions were machined from a butt-welded 2 mm thick Ti-6Al-4V sheets using a laser power of 2.1 kW, welding speed of 2.1 m/min and a positive focal position of +4 mm above the top surface of the workpiece and argon shielding gas at a coaxial flow rate of 25 L/min, a back protection flow rate of 20 l/min and a drag cover flow rate of 25 L/min. Digital images of the deforming surface of the specimens during deformation were captured every second for calculation of the

141 corresponding displacement and strain using the DIC technique. Figure 73 shows the resulting overall full field y strain distribution obtained from standard tensile testing at different times relative to the time to fracture of 20, 40, 60, 80, 95 and 99%. The legend for each stage was independently chosen to maximise contrast.

It was evident from a steep strain gradient along the loading direction as illustrated in Figure 73 that the local strain concentration was the maximum within the base metal, while only small deformations occurred in the FZ and the HAZ at the centre. All the tested weld specimens were over-matched and failed in the BM, indicating that the BM was the weakest region. As shown by the hardness and tensile testing measurements, the FZ and the HAZ were harder and stronger than the BM. It was discussed in Section 3 that the level of welding defects such as porosity and undercut observed in fibre laser welded Ti-6Al-4V welds were not critical enough to significantly affect the mechanical properties of the FZ and the HAZ. The great advantage of fibre laser welding Ti-6Al-4V compared to the results from the literature based on other laser sources such as CO2 and Nd:YAG lasers was that the tensile properties were not largely influenced by the welding defects so degradation in weld mechanical properties such as tensile strength and ductility was minimised [186,187,321–323].

Figure 73 Full field y strain distributions in loading direction for fibre laser welded Ti-6Al-4V showing the development of strain localisations at different fraction of the time to fracture

The global mechanical response of the welded specimen proved that welding increased strength and hardness in the FZ and the HAZ but decreased ductility in general. Still, the global tensile properties of the weld were similar to those of the BM in terms of elastic modulus, yield strength and ultimate tensile strength (UTS) except with a lower ductility. The joint efficiency defined as the ratio of the tensile strength of the weld to that of the BM, was approximately 100%, which indicated that the joint efficiency of the weld was as high as the BM. The local

142 constitutive stress-strain behaviour of different zones of the welded joint including the FZ, the HAZ and the BM were directly extracted from localised regions assuming a globally applied iso-stress level at all locations within the displacement data field. The local constitutive response in the FZ and the HAZ in Figure 74 was different to that in the BM and the global response of the specimen. The yield strength measured in the FZ was the highest with greater initial work hardening rate followed by the HAZ and the lowest in the BM. The higher hardness and local yield stress values in the FZ and the HAZ were attributed to the formation of martensite which increased strength at the cost of reduced ductility compared to equiaxed microstructure of the BM [324,325]. It was therefore concluded that the strengthening effect of the FZ outweighed the negative effect of the welding defects, provided that their size remained below the critical values. a) b) 3.06 1200

y = 0.0511x + 3.1224 1000 3.05

800 y = 0.0483x + 3.1216

600 3.04

Truestress (MPa) 400 Log Log true stress(MPa) HAZ 3.03 FZ 200 BM FZ BM (Eng) Global HAZ Global (Eng) 0 3.02 0.00 0.05 0.10 0.15 -1.9 -1.8 -1.7 -1.6 -1.5 -1.4 -1.3 True strain Log true strain c)

1200

1000

800 HAZ

600 FZ BM

True stressTrue(MPa) Global 400

200

0 0.00 0.02 0.04 0.06 0.08 0.10 True strain

143 Since far-field loading was limited by the strength of the weakest BM of the weld due to the nature of tensile testing transverse welded specimens which could accommodate strain preferentially to weaker regions, further straining to obtain full plastic deformations in the FZ and the HAZ was not possible, and therefore, plastic strain localisation and failure occurred outside the weld area, in the BM of the gauge length. As a result, the complete tensile response was obtained using the DIC strain mapping, only in the BM where the strain localisation occurred. In order to determine the constitutive behaviour of the FZ and the HAZ, the initial local stress and strain curves from the DIC were plotted to determine the power law relationship between the stress and the amount of plastic strain and predict the plastic proportion of the incomplete curves up to moderate values and estimate the ultimate tensile strength of both zones according to the Holloman’s power law work hardening model as shown in Equation 6 [326,327].

푛 σ = Kε푝 Equation (6) where σ is the stress, K is the strength index, 휀푝is the platic strain and n is the work hardening exponent. The calculated n was 0.0483 in the FZ, and 0.0511 in the HAZ. It was therefore, possible to evaluate the weld strength mismatch and plastic properties of different weld subzones. The calculated tensile strength of the FZ and the HAZ was higher than that of the BM so the weld strength was not determined by the FZ and the HAZ but by the properties of the BM. It was estimated based on the higher strength of the weld that the elongation to failure of the weldment is lower than that of the BM as martensite in the FZ and the HAZ shows poor ductility. Figure 74 shows the local y strain distribution in the loading direction measured across the welded joint using the DIC. It was found that the strain was the lowest in the FZ of around 3% and progressively increased with increasing distance away from the weld centreline, passing through the HAZ at around 5% and then finally reaching a maximum value in the BM. The opposite trend was observed with the micro-indentation hardness measurements, where the micro-hardness was the highest in the FZ of around 390 HV, slightly lower values of around 370-380 HV in the HAZ and the lowest in the BM of around 330-340 HV. It was clear that martensitic solid state phase transformation during welding had the effect of inherently increasing the hardness and local yield and tensile strength values in the FZ and the HAZ at the expense of ductility. For this reason, premature failure by strain localisation in the FZ or the HAZ could be avoided.

(ii) To verify the results obtained from tensile testing macro-tensile specimens, micro-tensile specimens were also tested under various welding conditions to examine whether or not the global tensile properties of Ti-6Al-4V fibre laser welds were affected by changing welding parameters. The main advantage of using the DIC for micro-tensile testing was that the spatial

144 resolution was only limited by optical magnification and the size of speckle patterns of appropriate spatial scale. By combining the DIC with an optical microscope, it was possible to obtain a close up view of the deformation within the weld during tensile testing as the weld occupied a larger proportion of the gauge length in micro-tensile specimens, and also achieve a more uniform plastic strain distribution over the gauge length. The detailed measurement results global constitutive behaviour under different welding conditions can be found in Appendix B. It was found as shown in Figure 75 that the global tensile properties of the welded specimens were similar to that of the BM and remained fairly constant over a wide range of laser powers, welding speeds and focal positions because all of them failed in the BM. Almost no variations in elastic modulus and elongation to failure were observed with changing welding parameters but tensile strength was affected by a small amount. It was obvious that the tensile strength values in the FZ and the HAZ were higher than that of the BM due to finer prior β grain structure and fine martensitic structure which exhibited greater resistance to deformation than the equiaxed α + β structure in the BM. As all the specimens failed in the BM, the tensile properties of these stronger zones were not the deciding factor for the global tensile properties but rather those of the weaker BM. However, the apparent variations in tensile strength with changing welding parameters as shown in Figure 75 was attributed to the change in weld shape for different welding conditions. Depending on the weld width of the weld seam, the ability to distribute localised plastic strain evolution more efficiently within the gauge length was different. Since the specimen thickness was only 0.4 mm, the effect of weld shape was only examined in terms of the weld width.

At lower laser powers below approximately 3 kW for the investigated welding speeds up to 5.0 m/min, tensile strength increased with increasing laser power because weld width became wider with increasing laser power and also weld shape became more uniform with flat edges, whereas, decreasing laser power led to a V shape with narrower root width, before extracting the specimens from the workpiece. Such trend was observed at welding speeds from 2.1 m/min to 5.0 m/min and laser powers from 1.4 kW to 2.6 kW. At higher laser powers in the range between 3.3 and 4.5 kW, any further increase in laser power above 3.3 kW resulted in lower tensile strength with increasing laser power.

As previously discussed in Section 3.3.1, for a similar set of welding conditions, the optimum weld quality was achieved at around 3.3 kW and any further increase in laser power made almost no change in weld width but at the same time produced more welding defects such as larger depth of undercuts. Therefore, increasing the laser power produced wider weld seams which improved the joint efficiency by spreading the load more uniformly over the specimen but above a certain level of laser power, the tensile strength was reduced due to greater influence of welding defects. Variations in welding speeds in the range between 2.1 and 3.3

145 m/min at a laser power of 2.5 kW resulted in a decrease in tensile strength of the specimens with increasing welding speed. It was also found that as the welding speed increased, the elongation to failure increased slightly.

a) 1200 V=2.1 m/min, f=+2 mm, Ar

1000

800

600

Truestress (MPa) 400 P=1.4 kW P=1.6 kW 200 P=1.8 kW P=2.0 kW 0 0.00 0.02 0.04 0.06 0.08 0.10 True strain b) 1200 P= 2.5 kW, f= +2 mm, Ar

1000

800

600

Truestress (MPa) 400 2.1 m/min 2.4 m/min 200 2.7 m/min 3.0 m/min 3.3 m/min 0 0.00 0.02 0.04 0.06 0.08 0.10 True strain c) 1200 P=2.1 kW, V=2.1 m/min, Ar

1000

800

600

Truestress (MPa) 400 f=0 mm f=+2 mm 200 f=+4 mm f=-2 mm f=-4 mm 0 0.00 0.02 0.04 0.06 0.08 0.10 True strain

Figure 75 Global tensile properties of Ti-6Al-4V fibre laser welds processed under different laser power, welding speed and focal position

146 As discussed above, increasing the welding speed led to a more refined martensitic microstructure in the FZ causing a steeper tensile properties gradient between the FZ and the BM, reduced weld width and a switch in weld shape from hourglass to V shape. A combination of all these effects led to deteriorated tensile strength but slightly improved elongation to failure with increasing welding speed due to increased joint microstructure heterogeneity.

The effect of changing focal position on tensile properties while maintaining the laser power and the welding speed constant, was such that changing the focal position influenced the power density of the incident laser beam on the workpiece and therefore, increasing the height of the focal position in any positive or negative directions led to a wider weld width which for the similar reasons as above increased the tensile strength of the specimens. It was important however, to note that all these observations were based on the analysis of the global tensile properties of the welded specimens and not the local tensile properties of different microstructural regions of the welded joint. In addition, even though certain amount of welding defects were observed in the specimens, their influence was not adverse enough to affect the overall mechanical properties and the failure location of the over-matched Ti-6Al-4V welds.

Figure 76 shows the DIC images of the micro-tensile specimens in Figure 75 welded under various laser powers at a constant welding speed of 3.0 m/min and focal position of +2 mm just before failure.

2.0 kW 2.2 kW 2.4 kW 2.6 kW Figure 76 Processed and raw DIC image data of fibre laser welded Ti-6Al-4V micro-tensile specimens as a function of laser power just before fracture and the equivalent chemically etched specimens without speckle patterns showing the location of fracture in the base metal

147 For these specimens, it was not possible to achieve good enough contrast using the etched microstructure so speckle patterns were applied for the DIC and another set of etched specimens were tested without the DIC for further analysis to observe the failure region more clearly. As it can be seen, the different colours indicate different strains prior to failure and the maximum localised strains close to 25% were observed outside the HAZ and within the BM off centre where fracture took place indicating that the BM was the weakest region as confirmed by the etched specimens without speckle patterns. The strain localisation was the minimum in the FZ where the martensitic microstructure was the strongest but also less ductile. An increase in the FZ width with increasing laser power was also observed in these images.

(iii) Figure 77 shows the strain maps of transverse and longitudinal T-joint specimens at different stages during tensile testing. Unlike the AA 2024-T3 longitudinal T-joint specimens, a larger radius of curvature at the neck was used for the Ti-6Al-4V longitudinal T-joint specimens as previously shown in Figure , in order to locate the fracture position within the gauge length. Otherwise, the initial specimen geometries led to fracture either at the weld run in or out positions due to weld strength overmatch. Even with such modification, failure occurred very close to the specimen transition region. In both types of specimens, the stiffener was not removed to examine its influence on the joint mechanical properties and to simulate the loading condition in a real fuselage and a result, a bending moment was induced due to asymmetry of the stiffeners with respect to the loading plane. It was found that the stress distribution in the transverse or the hoop stress specimens was the similar to that in the butt welded specimens in Figure 73, where strain localisation and the maximum plastic deformation occurred in the weaker BM with resultant necking and failure outside the fillet welded T-joint area. On the other hand, a different behaviour was observed for the longitudinal specimens where all zones deformed simultaneously with the same strain as shown by a uniform strain distribution perpendicular to the loading direction across the width of the specimen until failure occurred at the limiting strain. The strain was localised at the converging sections at either ends as well as at stiffener run in and run out locations. Failure occurred in one of these positions and stiffener separation suddenly occurred originating at the failure position of the base plate. The crack propagated along the base of the stiffener just above the welded joint and the stiffener separation progressed all the way to the end of the stiffener. The maximum local strain values just before failure were similar in both transverse and longitudinal stress specimens of around 40%.

148 a) Transverse T-joint

b) Longitudinal T-joint

Figure 77 Processed and raw DIC image data of fibre laser welded Ti-6Al-4V a) transverse and b) longitudinal T-joint tensile specimens at 50%, 75%, 99% and 100% of the time to failure

The strain distribution into the stiffener and the base plate for the stiffener tension test in Figure 78 showed that the strain concentration increased at the weld toe on the base plate side within the BM where the maximum constraint existed. The local strain around the welded joint or within the FZ at the centre of the specimens was very low with a value less than 5% throughout due to the local strengthening of the martensitic microstructure in the FZ, which led to higher resistance to deformation in the FZ. As a consequence, crack propagated along the weld toe and fracture took place on the base plate side outside the stiffener area while the welded joint remained intact. It was found that once a threshold load was exceeded, the specimen began to deform and eventually failed in the bent area around the weld toe.

Figure 78 Processed and raw DIC image data of fibre laser welded Ti-6Al-4V T-joint stiffener tension specimen at 50%, 75%, 99% and 100% of the time to failure

149 Figure 79 shows that the effective stiffness of the longitudinal T-joint specimens increased considerably to around 190 GPa, whereas, it decreased by only a limited amount in the transverse T-joint specimens, to around 107 GPa, compared to a value of 113 GPa in the unwelded BM specimens and 126 GPa in the butt welded specimens. Tensile strength of the longitudinal specimen also increased to almost 1250 MPa, whereas, elongation to failure decreased significantly to around 0.09 relative to the unwelded BM of 0.14. In the case of the transverse specimen, the elongation to failure slightly decreased to 0.13 but not as much as the longitudinal specimen, whereas, its tensile strength was almost identical to that of the unwelded BM, approximately 1100 MPa. Both the higher tensile strength and lower ductility values in the longitudinal T-joint specimens were attributed to the contribution of the inherent characteristics of the very fine martensitic microstructure in the FZ and partially in the HAZ during deformation which led to local strengthening of the specimen at the expense of ductility [321]. The peak stress measured in stiffener tension specimens on average was much lower than the values recorded in transverse and longitudinal stress specimens as only limited amount of plastic deformation occurred in these specimens and failure occurred via combined tension and bending loading. Initially, the maximum load was achieved while applying tension to the stiffener. When the load reached a critical value, any further increase in displacement caused a rapid drop in the applied load as the specimen started to deform by bending and then the load gradually increased until fracture. a) b) 18 450 1200 16 400

1000 14 350

12 300 800 10 250

600 8 200

Force (kN) Stress (MPa)

Truestress (MPa) 6 150 400 PM Butt 4 100 200 Fillet (Trans) 2 50 Fillet (Long) 0 0 0 0.00 0.03 0.06 0.09 0.12 0.15 0 1 2 3 4 5 True strain Displacement (mm)

4.4.3 Preliminary Physical Simulation of Ti-6Al-4V Weld Microstructure As mentioned previously, Ti-6Al-4V welds were over-matched due to martensitic phase transformation during the welding process with the weld metal strength exceeding that of the base metal, the maximum plastic strain localisation occurred in the unwelded region of the

150 gauge length with resultant necking and fracture in the weakest BM. In addition, the influence of welding defects on mechanical properties of fibre laser welded Ti-6Al-4V specimens was negligible. Therefore, it was not possible to obtain the full plastic deformations in the FZ and HAZ using the 2D DIC technique as the specimens failed in the BM. For this reason, in order to estimate of the full stress–strain curves of the FZ and HAZ, and compare with the experimentally determined curves at different cooling rates as shown in Figure 80, the incomplete plastic proportion were extrapolated using a power law strain hardening model as illustrated in Figure 74.

Complex microstructures within a very narrow width of the FZ and the HAZ produced due to steep temperature gradients during fibre laser beam welding, made it difficult to measure their individual tensile properties. In order to verify the above predictions, the Gleeble thermo- mechanical simulator was used to examine the microstructural evolution of Ti-6Al-4V as a function of cooling rate as shown in Figure 80. It was possible to simulate thermal cycles with various cooling rates and produce a homogenous single microstructure within a large volume of the testing specimen using the Gleeble [242] so the microstructure and properties of Ti-6Al- 4V for different cooling rates could be tested and compared to that of the FZ of welded specimens. a) b)

1000 20 (Output) 100 (Output) 1200 200 (Output) 300 (Output) 400 (Output) 800 20 (Input) 1000 100 (Input)

200 (Input) C)

300 (Input) 800 600 400 (Input)

600

400

True True stress(MPa) Temperature( 400 20 100 200 200 200 300 400 FZ 0 0 0 20 40 60 80 0.00 0.02 0.04 0.06 0.08 0.10 Time (s) True strain

The testing specimens were initially at the same state as the BM. As Figure 80 shows, thermal cycles were programmed into the Gleeble as input and the temperature history during testing was measured using thermocouples as output. Initially, attempts were made to simulate precisely the welding thermal cycles obtained from thermocouple measurements during actual welding trials. However, difficulties were encountered when simulating the non-isothermal welding thermal cycle, which involves both slow and rapid heating and cooling rates. As a

151 result of such limitations, the temperature profiles were simplified to consist of various isothermal cooling rates. It would be necessary to further develop this preliminary investigation in the future in order to accurately simulate the real welding thermal histories. The first stage of this preliminary experiment began by heating to above the β transus temperature of around 1000°C, at identical heating rates and then held at the maximum temperature for sufficient time to completely dissolve the hexagonal closed packed (HCP) α phase and undergo allotropic transformation to the body centre cubic (BCC) β phase. The next stage involved cooling at various rates from 20°Cs-1 to 400°Cs-1. Air quenching was used at lower cooling rates and water quenching was used when a high cooling rate was required. In the final stage, the specimens were loaded in tension at room temperature in vacuum condition at a constant strain rate of 0.001 s-1 to evaluate the mechanical behaviour of specimens with different microstructures resulting from different transformation kinetics on cooling.

The tensile testing results in Figure 80 showed that as the cooling rate increased the room temperature tensile strength also increased but the elongation to failure decreased. However, as expected the change in elastic modulus and yield strength with cooling rate was relatively small. The faster cooling rate was equivalent to increasing welding speed or decreasing laser power during welding which led to microstructure refinement in the FZ [137] and therefore, reduced the tensile strength and elongation. It was found that the elastic behaviour in the FZ of the welded specimen matched those of the simulated specimens cooled at a rate within the range of 300 and 400°Cs-1, whereas, the predicted plastic behaviour was lower but instead, closer to those of 100 to 200°Cs-1. It may be due to under prediction of the amount of hardening in the actual FZ of the weld during deformation. The results also indicated that ductility considerably decreased with increasing cooling rate.

Figure 81 shows that different microstructures and phase morphologies were formed depending on the temperature history and the cooling rate. Slow cooling through the β transus at 20°Cs-1 led to a diffusion controlled transformation of the β phase into a mixture of α + β phases. The allotriomorphic α phase nucleated at prior columnar β grain boundaries, and also formed Widmanstätten primary α platelets by nucleation and growth in a colony morphology into β grains from the grain boundary α or from the prior β grain boundary, with the colony density decreasing with increased cooling rates [282,328]. Under moderate cooling from above the β transus above 20°Cs-1 and below 400°Cs-1, different transformation kinetics on cooling were observed and diffusional growth of primary α was restricted, where increasing fractions of acicular α (but not lath α also known as massive α) and martensitic α´ compared to primary α phase formed from the β phase. The acicular alpha phase preferentially formed at β grain boundaries with high dislocation density by a diffusion controlled transformation,

152 while α´ formed at and intragranularly within the prior β boundaries as orthogonally oriented high aspect ratio needles or plates by diffusionless transformation [282,329].

At the highest cooling rate of 400°Cs-1, it was found that the alpha transformation was completely suppressed and quenching from the β phase led to the formation of mostly α´ oriented orthogonally and a small fraction of acicular α. Martensitic solid state phase

153 transformation only involved a change in crystal lattice structure but no rearrangements of atoms. The transformation according to Cho [330] is very fast and only takes about 10- 7 seconds to complete the transition in a single grain so the propagation process is difficult to observe. Ahmed and Rack [282] determined in their investigation that the martensitic transformation in Ti-6Al-4V forms at cooling rates greater than 410 Ks-1, whereas, Gil et al. [331] observed a fully martensitic microstructure at cooling rates of only 5.1 Ks-1. However, according to the results shown in Figure 80 and 81, it was concluded that the cooling rate of 410 Ks-1 was more appropriate for this investigation and the cooling rate of 5.1 Ks-1 was too low. Hence, the microstructure of the FZ obtained from the actual weld cross-section as shown by the last micrograph in Figure 81, was similar to those resulting from a water quench process at fast cooling rates of around 300 Cs-1 from above the β transus temperature. Even though the ductility of the martensitic microstructure was lower than that of the BM, its strength was significantly higher. Therefore, as long as the ductility and toughness of the FZ is not too low, strength over-matching can be beneficial [11].

4.5 Conclusions

4.5.1 AA 2024-T3 Softening in the FZ and over-aging in the HAZ were caused by the dissolution or coarsening of the strengthening precipitates during welding where very high temperatures as well as rapid heating and cooling rates were experienced. The degree of softening was affected by the amount of heat input. The reduction in hardness in the FZ and the HAZ was maximised at higher heat inputs.

AA 2024-T3 welds were under-matched with the maximum joint efficiency of around 85%. The weaker strength of the FZ deteriorated the plastic straining capacity of the welded specimens due to increased stress concentration and constraints in the FZ which confined plasticity development within the weld. As a result, all specimens fractured in the weld, either in the FZ when the weld width was sufficiently wide or in the HAZ where the property gradient was the greatest when the weld width was relatively narrower. Welding only had a small influence on the elastic modulus and yield strength whereas, tensile strength and ductility were considerably reduced due to the presence of defects compared to the unwelded BM.

The specimens which showed poor weld mechanical properties failed in a mixed mode of brittle and ductile failure and contained micro porosities and hot cracks, whereas, the specimens with optimised welding parameters failed in a ductile mode and also welding defects were significantly reduced.

Increasing laser power led to higher ultimate tensile strength and ductility but lower yield strength due to increased heat input. On the other hand, increasing welding speed led to lower

154 tensile strength and ductility but higher yield strength due to the formation of finer dendritic structure and grain sizes in the FZ. Increasing or decreasing focal position relative to the top surface of the workpiece also resulted in wider weld width which increased ultimate tensile strength and ductility but lowered yield strength.

Longitudinal T-joint specimens showed better performance compared to butt welded due to contribution from the high stiffness of the welded stiffener, whereas, lower than the longitudinal T-joint specimens but were above the butt welded specimens for transverse T-joint specimens due to less influence of weld seam on plasticity development within the gauge length and strain localisation from geometrical notch effect.

4.5.2 Ti-6Al-4V The weld exhibited the largest hardness values due to the formation of fine needle like martensitic microstructure at fast cooling rates. The hardness in the HAZ gradually dropped with increasing distance from the weld centre due to a progressive drop in the martensite content. The hardness in the FZ decreased with increasing heat input which was achieved either by increasing laser power or decreasing welding speed. Changing focal position on the other hand, had only a small influence on the hardness variations within the FZ.

For all the welding conditions examined, the weld qualities were sufficiently good such that the influence of welding defects on weld mechanical properties were negligible. The local plastic deformation in the FZ was the minimum and so all specimens failed in the BM.

The global tensile properties of the welded specimens were limited by the properties of the BM and the joint efficiencies were close to 100% due to local strengthening effect of the martensite at the expense of reduced ductility. For this reason, the joint mechanical properties remained nearly constant over a wide range of welding parameters and no significant degradation in both the tensile strength and ductility was observed within the investigated process parameter window.

The mechanical behaviour of transverse T-joint specimens was similar to that of butt welded specimens, whereas, a significantly higher tensile strength but also a considerable drop in elongation to failure were observed in longitudinal T-joint specimens due to simultaneous deformation of both the BM and the stronger weld consisting of martensite. Stiffener tension testing produced similar results where the failure occurred in the BM due to the added strength in the welded joint from the martensite in the FZ and the HAZ and therefore, stiffener and base plate separation did not occur.

155 5 EXPERIMENTAL MEASUREMENT OF WELDING RESIDUAL STRESSES AND DISTORTIONS IN AA 2024-T3 AND Ti-6Al-4V

5.1 Introduction Welding thin section sheet materials generates considerable amount of residual stress mainly in the welding direction which often has large tensile stress in vicinity of the weld at room temperature, balanced by lower compressive stress in the rest of the workpiece, and also relatively low stresses in the perpendicular and normal directions. Residual stresses are self- equilibrating stresses caused by incompatible internal strain associated with plastic deformation, metallurgical transformations and solidification of the weld metal, which continue to exist even in the absence of external forces [332]. The non-uniform thermal expansion and contraction of the weld metal and adjacent base metal due to localised transient heat and strongly non-linear temperature fields [333] during the welding heating and cooling cycle induce plastic deformation and permanent volumetric thermal stresses cause welding induced distortions in thin sheets which are inherently not very stiff to resist the compressive shrinkage forces [333,334]. Depending upon the distribution of the shrinkage forces, the geometry and material characteristics of the welded structure, various types of distortion, such as bending, cambering, rotation and buckling may occur. In addition, the region far from the weld, less affected by the welding heat and the clamping system, act as restraints on the weld area and contribute to thermal distortion problems. Although distortions occur by the same mechanism as residual stresses they are opposite in tendency where high residual stresses lead to lower distortion and vice versa.

Figure 82 Evolution of temperature and stress in the welding direction during welding (figure taken from [335])

156 Figure 82 illustrates the development of the temperature field and residual stress profile in the welding direction during welding. Along section A-A ahead of the heat source, there is no temperature change and so the stress is zero. Along section B-B crossing the heat source, there is a large temperature change which leads to almost no stress in the weld pool since the weld pool is unable to support any loads, small compressive stresses around the weld pool due to thermal expansion in the weld pool are constrained by tensile stresses in the base metal [336]. Along section C-C behind the heat source, the temperature distribution becomes less steep and tensile stresses start to develop as the weld cools and shrinks at the weld centreline and compressive stresses in the base metal due to thermal misfit. Further away from the heat source along section D-D, the temperature change profile becomes uniform and close to zero when the weld has cooled to room temperature and the stress distribution remains the same but higher tensile and compressive stresses are produced in the corresponding regions and eventually become the residual stress distribution.

The yield strength is a function of temperature where it decreases with increasing temperature so the yield strength may be reached or even exceed on heating and result in plastic deformation and strain beyond this point, whereas, on cooling the yield strength increases as the temperature drops and the stress becomes elastic. The relative yield strengths of weld metal and base metal have an influence on the magnitude and distribution of residual stresses. For example, solid state phase transformation during solidification in Ti-6Al-4V, softening of the weld due to dissolution of the strengthening precipitates in heat treatable AA 2024-T3 and dilution between filler metal and parent metal when welding with filler metal can change the weld metal tensile properties [335]. Therefore, it is important to obtain the exact knowledge of the mechanical properties of the weld metal for each combination of welding parameters as studied in Chapter 4 [337]. In addition, welding parameters also have a strong influence on residual stress, where the temperature distribution, peak temperature and cooling rates in the weld are affected and the amount of energy absorption in the workpiece depends on the material, joint geometry, the type of heat source including the wavelength of the laser beam and the welding parameters [332,338,339]

Fibre laser welding leads to less distortion and residual stress than conventional fusion welding processes due to its lower overall heat input. Nevertheless, some residual stresses are formed which can have either positive or negative effects on integrity, fracture toughness, load capacity, stress corrosion resistance, fatigue life and fatigue crack initiation and propagation of the welded component under cyclic loading [336,338]. Tensile residual stresses are generally detrimental since they lead to brittle failure or accelerated crack growth near the weld region and increase the rate of damage by fatigue, whereas, compressive residual stresses can improve fatigue life by suppressing crack growth along the weld but may

157 also lead to buckling [334]. Distortion can also pose serious problems since the shape of the final component is affected. The main distortion mode in welded thin sheets and stiffened panels is out-of-plane distortion caused by angular change and cambering along the weld centreline. However, in thin sheets, the angular distortion is not significant due to the relatively small temperature gradient through thickness [340].

A good understanding of the welding process and information about residual stresses and distortions in welded components are of great interests for quality control and improvement of structural performance of integral structures so that in service failures can be avoided [333]. It is therefore, important to investigate the effect of residual stress and distortion on the structural performance and mechanical properties of the welded joints such as the tensile, fracture and fatigue behaviour and reduce them to acceptable levels by using optimised welding parameters, joint geometry and well characterised welding procedures [333].

Determination of residual stresses and distortions by experimental measurements is complex, expensive, time consuming and requires significant amount of resources. Also, subsurface measurements in the past were limited to destructive and indirect methods until more advanced non-destructive neutron diffraction and synchrotron techniques became available [338]. Moreover, it is difficult to obtain detailed full field temperature and residual stress maps of the welded component. Increasing use of numerical simulation of welding in recent years using finite element method has allowed a more resource effective way of estimating transient thermal stress, residual stresses and distortions in comparison to experimental methods. However, predictions from weld modelling still require calibrations based on precise experimental measurement of welding thermal cycle, and constitutive behaviour of the weld metal, heat affected zone and base metal [332]. In this way, it is possible to understand the welding process and perform parametric studies to identify optimal welding parameters and control residual stresses and distortions in order to minimise the use of post weld stress relieving procedures. It is much easier to produce structures without distortion than to reduce it after welding through post weld mechanical stretching [340].

The methods of controlling welding distortions involve selection of appropriate restraint conditions such as supporting material and fixtures [335]. Similarly, while it is possible to relieve residual stresses in heat treatable alloys by post weld heat treatment, it is not practical in the case of large structural applications to perform a PWHT on the entire structure due to difficulties in solutionising the material at high temperatures and then quenching without considerable deformations as well as problems associated with overaging in the HAZ for AA 2024-T3. Applying the PWHT locally is not recommended as it will also generate additional temperature gradient and result in the same overaged structure. For this reason, a lot of welded structures are designed to operate in as welded condition without PWHT so they must

158 be designed in such a way to minimise residual stresses [335]. For the purpose of this investigation, macroscopic residual stresses were of the most relevance.

5.2 Experimental Measurement of Welding Distortions Angular distortion and cambering out-of-plane displacements due to thermal expansion and contraction during welding in the welded plates and T-joints were measured with a Nikon LK G-90C coordinate measuring machine (CMM) and Camio 6.3 software as shown in Figure 83. CMM is a high precision instrument for measuring the physical geometrical characteristics such as surface contour of an object. A motorised automated probe head with an electronic touch trigger probe (TP200) was programmed using the software to collect a series of point measurements along three lines in both x and y directions to the weld centreline with 10 mm spacing between points, by quickly pecking the surface over the entire specimen.

Figure 83 Experimental setup for measuring the out-of-plane displacement after welding using coordinate measuring machine 5.3 Experimental Measurement of Welding Residual Stresses There are various methods available for measuring residual stresses (RS) such as nondestructive X-ray and neutron diffractions techniques used in this investigation [341]. A comparative description of the available techniques for residual stress measurements is given by Wither and Bhadeshia [342], and a review of the neutron diffraction technique and its applicability by Allen et al. [343]. The purpose of using these techniques to measure welding residual stresses on small scale components is to experimentally validate numerical models so that its outputs are representative for the real welded structures, and to analyse the influence of welding parameters on the magnitude and distribution of residual stresses. It is therefore, necessary to make the weld model predictions as accurate and realistic as possible so that largely conservative assumptions such as simple yield strength level residual stress profiles defined in BS 7910, R6 and FITNET can be reduced, and more efficiently meet the in-

159 service operating requirements and ensure structural integrity of a welded component. Staron et al. [344] studied residual stresses in CO2 laser welded 3.2 and 6.0 mm thick AA6056 butt joints using neutron and high-energy X-ray diffraction and showed that the residual stress level, in fact reached only around 70% of the material’s yield strength at room temperature in the thinner sheet, whereas, nearly 90% in the thicker sheet, due to higher heat input required for welding a thicker material. A study by Ivetic et al. [345], who conducted residual stress measurements along the thickness of welded panels via electric strain gauge method, concluded that it is possible to assume a constant distribution of the stress through thickness, where thicknesses are up to around 25.4mm.

The diffraction technique measures lattice spacings in crystalline materials which have randomly orientated grains. The wavelengths of X-ray and neutron are of similar order of magnitude as the lattice spacings so the radiations can penetrate the crystal lattice and be reflected from successive atomic planes leading to diffraction [346]. The crystal lattice in each grain is used as an atomic strain gauge. When a given material is elastically deformed, the crystal lattice is distorted and results in changes in the lattice spacings, whereas, plastic deformation only causes slipping between lattice planes due to dislocation motion rather than affecting the lattice spacings [341]. Relative shift in the lattice spacings is used to determine the elastic residual strain in the material using Bragg’s Law which defines the condition for diffraction as shown in Equation 7.

2dhkl sin hkl   Equation (7) where λ is the wavelength of the radiation, dhkl is the lattice plane spacing between selected lattice planes in a crystalline material and θhkl is the Bragg angle at which the radiation is coherently and elastically scattered for a particular crystallographic plane. Either λ or dhkl is measured while keeping one of them constant, depending on the measurement instrument so that the lattice spacings can be calculated using the Bragg’s Law. Therefore, in order to determine the elastic strains, the stress free lattice spacing, d0 and the corresponding diffraction angle, θ0 must be known or determined separately. Accurate information on d0 is more important for the neutron diffraction technique than the X-ray diffraction technique as the X-ray is basically used for near surface effects where the stress component in the thickness direction is assumed to be zero. Such an assumption allows the calculation of d0 simply from the in-plane stress components. On the other hand, in the case where the out-of-plane stress component is large, then the d0 has to be determined separately, which can be difficult in the interior of a material.

The elastic strain can be calculated by the differentiated form of Equation 7, using Equation 8 [341]. The determined strain is an average elastic lattice strain over a sampled gauge volume

160 defined by slits or collimators and formed by the intersection of the incident and diffracted beams rather than that of atomic scale [347].

ddhklhkl 0, hklhklhklhkl  0,0, Cot Equation (8) d0,hkl

The direct stresses σxx, σyy and σzz can then be calculated from the linear elastic properties of the material and the measured residual elastic strain in the relevant directions as shown in Equation 9.

Ehkl hkl hkl hkl  xx  1 hkl  xx  hkl  yy   zz  1 hkl 1 2 hkl 

Ehkl hkl hkl hkl Equation (9)  yy  1 hkl  yy  hkl  xx   zz  1 hkl 1 2 hkl 

Ehkl hkl hkl hkl  zz  1 hkl  zz  hkl  xx   yy  1 hkl 1 2 hkl  where Ehkl and vh k l are the elastic modulus and Poisson’s ratio, respectively of a specific crystallographic plane. The diffraction peak specific elastic constants for texture free materials are listed in Table 10 for aluminium and Table 11 for titanium [348]. The values for aluminium are based on the Kröner model which assumes the stress and strain to vary between the grains in a crystalline material [348].

Table 10 hkl specific E and ν of aluminium following the Kröner Model [349]

{200} {311} {420} {531} {220} {422} {331} {111} Ehkl (GPa) 67.6 70.2 70.3 71.2 71.9 71.9 72.3 73.4 νhkl 0.35 0.35 0.35 0.35 0.34 0.34 0.34 0.34

Table 11 hkl specific E and ν of titanium [350] {103} {200} {112} {201} Ehkl (GPa) 121 110 113 110 νhkl 0.31 0.33 0.32 0.33

The reflections which are only weakly affected by residual intergranular stresses are recommended for measuring the lattice strain as listed in Table 12. A reflection that is affected by intergranular stresses cause the elastic lattice strain for that particular reflection to differ from the true macroscopic elastic strain and lead to non-linear lattice strain versus applied stress response due to plastic anisotropy [351]. Therefore, by choosing a reflection that is insensitive to plastic deformation, the influence of intergranular stresses can be avoided. The measured lattice spacings can also be affected by changes in chemical composition such as dissolution of strengthening precipitates during welding in AA 2024-T3, phase transformations and the presence of texture due to reorientation of grains.

161 Table 12 Lattice planes weakly and strongly affected by intergranular strains [351]

Material Weak Strong FCC (Al) {311}, {422}, {220} {200} HCP (Ti) {102}, {103} {002}, {100}, {110}

5.3.1 X-ray Diffraction X-ray is a form of electromagnetic radiation with wavelengths ranging from around 0.01 to 10 nm that is produced when accelerated electrons collide with a target material and interact [352]. A diffraction equipped with monochromatic filter and Cu K-α radiation source with a wavelength of 0.154 nm and 0.075 mm attenuation length was used as shown in Figure 84.

Figure 84 X-ray diffractometer equipped with a copper X-ray source

The X-ray diffraction technique is based on the constructive interference (diffraction) of X-rays incident on the lattice planes of crystals oriented in such a way to fulfil Bragg’s Law as shown in Equation 7 [352]. It is primarily used or even limited to measurement of near surface strains in crystalline materials to a maximum depth of around 0.05 mm [341]. Subsurface measurement requires destructive electrolytic polishing to remove successive layers or the use of other techniques [345]. Since the penetration depth is limited to a very thin surface layer, a biaxial stress state assumption is made where the out-of-plane stress is assumed to be zero, which makes it possible to use the sin2 ψ method to evaluate the in-plane stress state as a function of distance from the weld centreline. Since the wavelength of the X-ray is already known, the interplanar spacing or d spacing can be calculated by measuring the reflected beam intensity along a goniometer circle at a series of angles using a counter and receiving slit and finding the position of maximum beam intensity. The advantage of using the sin2 ψ method is that stress free reference do is not required. According to Belassel et al. [353] only elastic strains are measured when using this technique, which are dependent upon the applied load and plastic strain incompatibilities among grains and layers. In addition, it is necessary to calibrate the X-ray elastic constants either experimentally or theoretically as discussed by Belassel et al. [353] in order to account for the differences between the strains measured using

162 the X-ray diffraction technique and the macroscopic mechanical strains measured using strain gauge method.

5.3.2 Neutron Diffraction The neutron diffraction technique is identical in principle to the X-ray diffraction technique. It is based on the measurement of interplanar lattice spacings. The main difference is the penetrability of neutrons which unlike X-rays according to Pirling et al. [354] can penetrate deeply up to 10 cm in most crystalline materials even for heavy elements with a minimum sampled volume of 1 mm3. This means that neutron diffraction can be used to obtain information on residual stresses in the interior of a component. It also has the ability to readily adjust its spatial resolution using slits or collimators and resolve strain gradients within the local volume [347]. In some cases, an elongated matchstick shape gauge volume aligned parallel to a direction in which the strain gradient is small is used to maximise the data acquisition rate while maintaining a high spatial resolution in the direction of large strain gradients [351].

There are two main types of neutron diffraction techniques which depend upon on the neutron source. One is based on constant wavelength instrument while the other is based on variable wavelength instrument. In reactor based diffractometers such as SALSA and E3, a continuous monochromatic single wavelength neutron beam is produced from a polychromatic neutron beam by using a monochromator [355]. The beam direction and size is controlled by slits or collimators made of cadmium and the sampled volume known as the gauge volume is defined by the intersection of the incident and the scattered beam entering the detector. Since the incident wavelength of the diffracting neutrons is known, the lattice spacings of a specimen can be determined by measuring changes in the peak diffraction angle of a single diffraction peak as shown in Figure 85. Elastic strains can then be calculated using Equation 8. Karadge et al. [356] compared the residual stress measurements determined from SALSA using a single peak analysis to those conducted on ENGIN-X using multi-peak Rietveld analysis and found that the results showed excellent agreement. The largest stress differences of around 80 MPa in nickel base super alloy welds were identified at the weld line, which was attributed to alignment errors rather than principle differences between the two techniques.

163

Figure 85 Diffraction spectrum of fibre laser welded Ti-6Al-4V from reactor based diffractometer E3

In time pulsed spallation based diffractometers also known as time of flight instruments, neutron beams of varying wavelengths and velocity are produced in pulses via spallation from a heavy metal target such as tungsten or tantalum [355] on impact with a beam of high energy accelerated protons from a synchrotron source. The velocity of the neutron is used to determine its wavelength by measuring the time of flight over a known distance as shown in Equation 10.

hthkl   Equation (10) mn L

−34 where h is the Planck’s constant (6.6261×10 Js), mn is the neutron rest mass

−27 (1.6749×10 kg) [355], L is the total flight path distance and thkl is the time of flight of the hkl peak centre. Each pulse of neutron beam is polychromatic in wavelength so the whole diffraction pattern of Bragg peak intensity as a function of time of flight or wavelength can be obtained. Detectors are placed at a constant Bragg angle of 90° to the incident pulsed beam to collect neutrons diffracted from the specimen over many pulses. On a spallation source time of flight instrument, it is possible to detect scattering of neutrons simultaneously from both sides at 90° of the incident beam due to the absence of resolution focusing effect that is observed in continuous beam instruments [355]. The size of the incident beam on ENGIN-X is controlled using boron carbide slits [355]. Equation 10 can then be related to the Bragg’s Law to calculate the lattice spacing as shown in Equation 11.

hkl ht dhkl   Equation (11) 2sin hkl 2mn Lsin hkl

As the scattering angle is fixed, Equation 11 can be simplified to calculate the strain as given by Equation 12.

164 0 0 dhkl  dhkl thkl  thkl  hkl  0  0 Equation (12) dhkl thkl where thkl and thkl0 are the measured and reference time of flight for the hkl peak, respectively. Instead of conducting a single peak fit with Gaussian distribution on each individual peak, it is more common to fit an average lattice spacing to the whole diffraction pattern using the Rietveld (or Pawley-Rietveld) technique [357] as shown in Figure 86. The strain calculated using this method has been found to give a value close to the macroscopic strain in both elastic and plastic region so the macroscopic bulk elastic constants can be used to determine the stress [358].

Figure 86 Diffraction spectrum of fibre laser welded AA 2024-T3 from pulsed spallation source ENGIN-X fitted with Rietveld refinements

Various methods can be used to determine the reference lattice spacing. An ideal method is to use a reference stress free sample that is electrical discharge machined (EDM) in the form of a comb or coupons from an identical specimen to that being measured. Since the same intergranular strains exist in both the reference sample and the measured sample, they will cancel each other out. At the same time, the macroscopic residual stresses in the reference comb are relaxed. By comparing the d0 measured in the stress-free sample at the same locations in each direction as those measured in the sample, the variation in stress free lattice spacings can be taken into account and therefore, the correct absolute elastic strains can be obtained. Another method is to use a far field d0 value in the sample as a reference where the sample is unaffected by the welding process and the residual stress is expected to be very small. In this case however, it is important to apply stress and momentum balance across any plane cross-section of the sample as shown in Equation 13 and Equation 14 to verify that the far field region is indeed in a low stress state. If not, the reference value is calculated by iteration until a value is found in which balance occurs [359].

165 ∫ 휎푑퐴 = 0 Equation (13)

∫ 푑푀 = 0 Equation (14)

However, when using a global strain free reference value, the effect of intergranular stresses is not taken into account. Also, there may be local chemical compositional changes, phase transformation and microstructural gradient in welded samples which affect the reference value so the method can be inappropriate for measuring welding residual stresses. Another approach for relatively thin sheets in the absence of reference stress free sample, is to assume a plane stress condition and the constitutive equations reduces to those as shown in Equation 15.

Ehkl hkl hkl  xx  2  xx  hkl yy  1 hkl 

Ehkl hkl hkl Equation (15)  yy  2  yy  hkl xx  1 hkl 

 zz  0 ______

The reference lattice spacing and Bragg angle in a stress free sample can be calculated assuming plane stress using the follow relationships Equation 16 for d0 and Equation 17 for θ0.

0  1 d  (d  d )  d Equation (16) 1 x y 1 z

0  1   (  )   Equation (17) 1 x y 1 z

Residual stress distributions were measured by X-ray and neutron diffraction techniques for butt welded AA 2024-T3 and Ti-6Al-4V, and T-joint fillet welded AA 2024-T3 specimens as shown by the measurement positions in Figure 87 and Figure 88. Measurements were taken at around 30 points across the plates’ mid thickness and mid length at 1 to 2 mm intervals near the weld and increasing increment size away from the weld. The small increments near the weld was necessary to capture the large stress gradients in the weld region. It was expected that the distribution is symmetric around the weld centreline so it was valid to assume symmetry in the planes parallel and transverse to the welding direction and also assume that the values measured along the x and y directions are the principal stresses [345]. In addition, for the stiffened panels, measurements were also made at around 5 points along the height of the stiffener. Measurements were taken in three mutually orthogonal directions i.e. along the weld (y), transverse to the welding direction and normal to the sheet’s surface.

166 Measurement Positions

500 mm

250 mm

Y (Longitudinal)

400 mm

1.5 mm 3 mm X (Transverse) 400 mm Figure 87 Residual stress measurement positions for butt welded AA 2024-T3 and Ti-6Al-4V sheets

Measurement Measurement Fillet weld Stiffener Positions Positions Stiffener Fillet weld 25 mm

150 mm

300 mm 221 mm 400 mm 380 mm Z (Normal)

Y (Longitudinal) Measurement Positions 3.2 mm Y (Longitudinal) X (Transverse)

25 mm

X (Transverse) 3.2 mm

221mm Figure 88 Residual stress measurement positions for T-joint fillet welded AA 2024-T3 thin sheets with a single stiffener and triple stiffeners

For X-ray diffraction measurements, a diffractometer equipped with Cu K-α radiation source with the X-ray tube operating at 20 kV and 4 mA target current was used. The x and y residual stresses distributions were measured at a depth of 30 μm from top surface of samples and perpendicular to the welding direction. Neutron diffraction measurements were taken using three different instruments. The ENGIN-X pulsed spallation source at ISIS was used to measure two butt-welded AA 2024-T3 specimens, W99 and W102 as shown in Figure 90. A gauge volume of 2 × 2 × 2 mm3 was used to measure the y strains parallel to the welding

167 direction and 2 × 20 × 2 mm3 in the x and z directions. Stress free parameters were measured also using a gauge volume of 2 × 2 × 2 mm3 under identical conditions at every 2 mm interval from the weld centreline and the dimensions of the combs are given in Figure 89. The strains were determined by Rietveld refinement as a function of x distance from the weld centreline.

Figure 89 Dimensions of electrical discharge machined (EDM) stress free reference combs

168 The remaining butt welded AA 2024-T3 specimen W47 and Ti-6Al-4V specimen were measured using the E3 continuous reactor source at HZB. A gauge volume of 2 × 1 × 2 mm3 for AA 2024-T3 and 2 × 2 × 2 mm3 for Ti-6Al-4V was used to measure y strains parallel to the welding direction, whereas, a 2 × 19.3 × 2 mm3 matchstick shape gauge volume in the x and z directions for both materials. The elastic strain response of both {222} and {311} reflections were measured for the AA 2024-T3 specimen. On the other hand, measurements in the Ti- 6Al-4V specimens were proven problematic as very weak diffraction patterns were observed in certain orientations. While three peaks {103}, {112} and {201} were visible in the x direction, the {103} peak was not observed in the y direction and the {201} peak was very weak in the z direction. The likely cause of this phenomenon was texture in the weld material created during the solidification phase of the welding procedure which is frequently observed in hexagonal close packed (HCP) materials like titanium.

Figure 91 T-joint fillet welded AA 2024-T3 specimens mounted on the hexapod platform of the SALSA neutron diffractometer at ILL

T-joint fillet welded AA 2024-T3 stiffened panels, one with a single stiffener and the other with three stiffeners, were measured using the SALSA continuous reactor source at ILL as shown in Figure 91. SALSA provides high neutron flux and it is equipped with a 6 degree of freedom hexapod platform consisting of six hydraulic struts which allow the device to move in all six

169 coordinates (translation and rotation). For this reason, the stiffened panels which have considerable angular distortions were measured using the SALSA to precisely position the measurement points. A 2 x 0.6 x 2 mm3 gauge volume was used to make the residual stress measurements in the y direction, whereas, in the other two directions, a 2 x 10 x 2 mm3 elongated matchstick shape gauge volume with its long axis aligned parallel to either x or z direction was used. In this case, collimators were used instead of slits to define the size of the incident beam. The elastic strain response of the {222} lattice plane was measured in these tests because the response from the {311} reflection did not exist or was very weak. The samples were set up with an angle, 2θ, of 80.502° between the incident and scattering beam. Stress free measurements were made from two coupons, one machined from the weld and the other from the base metal as shown in Figure 89.

In addition, far field measurements were made at the edge of the sample, far from the weld line which was expected to be stress free. Since the d0 values used were only taken from two positions, one from the weld and the other from the parent, it may be necessary to measure additional points using combs to measure the d0 values in the same measurement points in the sample to obtain the correct elastic lattice strains. The reference d-spacing is likely to vary throughout the sample as different regions of the material have experienced different thermal cycles during welding. The possibility of variation in d-spacing due to local compositional changes from welding with filler wire were also cancelled out using combs as shown in Figure 89.

5.4 Results and Discussion on AA 2024-T3 (i) Surface residual stress measurements were taken using a low energy X-ray diffraction (XRD) technique on the top surface of butt welded AA 2024-T3 sheets, W47, W99 and W102, at a depth of around 10 μm from the surface. Figure 92 shows the bi-axial residual stresses, x and y to the welding direction, measured using the XRD compared to numerically calculated stresses.

It was assumed that the residual stress distributions are symmetrical about the weld centre line and therefore, only half of the sheet was measured. Both the experimentally measured and simulated x residual stresses were found to be within the maximum range of less than ±60 MPa for all three welding conditions, which when compared to the yield strength of 345 MPa were negligible. No obvious trend in measured x stress distributions were observed for all three specimens and appeared to be randomly scattered, whereas, numerical distributions showed a small peak at the centre and troughs in the adjacent regions which eventually levelled out. It was therefore concluded that the stresses in the x direction were relatively small in this thin sheet.

170 ai) aii) 80 250 σx σy 60 200

40 150

20 100 XRD 0 50 FEA -20

Transverse(MPa) stress

Longitudinal Longitudinal stress (MPa) Stress Stress (MPa) -40 Stress (MPa) XRD -60 -50 FEA -80 -100 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 DistanceDistance from from weld weld centre center (mm) (mm) DistanceDistance from from weld weld centre center (mm) (mm) bi) bii) 80 250 σx σy 60 200

40 150

20 100 XRD 0 50 FEA

-20 Stress Stress (MPa)

Stress Stress (MPa) 0 Transverse stress(MPa) -40 Longitudinal stress (MPa) XRD -60 -50 FEA -80 -100 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld center (mm) Distance from weld center (mm) ci) cii) 80 250 σx σy 60 200

40 150

20 100 XRD 0 50 FEA

-20 Stress Stress (MPa)

Stress Stress (MPa) 0 Transverse(MPa) stress -40 Longitudinal stress(MPa) XRD -60 -50 FEA -80 -100 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld center (mm) Distance from weld center (mm)

171 FE predictions on y residual stresses were in good agreement with experimental results. The predicted stress distributions showed large variations only up to around 5 mm away from the weld centre line and then levelled off to almost zero. The y stresses were found to be significant even very close to the surface according to both simulated and experimental results. The simulated and experimental y stresses had approximately the same magnitude and distribution, reaching around 70% of the room temperature yield strength around the weld. Peak tensile y stresses of around 220-240 MPa were predicted and observed in a very narrow region adjacent to the weld at approximately 3-4 mm from the weld centre which quickly dropped to below 0 MPa and became weakly compressive at distances greater than 5 mm from the weld centre. The experimental results indicated larger compressive stresses away from the weld at distances greater than around 5 mm from the weld centre than the predicted stresses, with the maximum difference of around 60 MPa. However, uncertainty in measurements as shown by the error bars meant that such differences were acceptable.

The stress distributions and magnitudes were similar among all three but minor differences were observed at the weld centre and in the width of the tensile region which corresponded to the width of fusion boundaries. More importantly, softening or heat affected weakening effect at the middle position in the weld, showing a drop at the centre was captured both in the experimental and numerical results, where a small dip was identified at the weld centre as shown in Figure 92. It was found that the lowest stress magnitude was measured and calculated in the specimen welded with filler metal, W102, which meant that this specimen experienced softening the most. The measured value of around 70 MPa was lower than the calculated minimum value of around 120 MPa at the weld centre. Similarly, the measured value at the weld centre of around 100 MPa for W47 was slightly lower than the calculated value of around 130 MPa. On the other hand, the largest value was measured in W99 of around 170 MPa, which was in good agreement with the calculated value of also around 170 MPa. The peak tensile y stresses were also the greatest in W99 of around 230 MPa, which was welded using a lower laser power and slower welding speed, compared to W47 and W102. Nevertheless, it was concluded that the predicted stress magnitudes and distributions illustrated in Figure 92 were in good agreement with the measured results. It seems too conservative according to these results to assume that yield magnitude residual stresses occur in AA 2024-T3 laser welds. Peak residual stresses as high as the yield strength are likely to occur when the thermal contraction strain becomes larger than the yield strain. Often in steels, the thermal strain can reach more than the yield strain and result in yield magnitude residual stresses. However, in the case of AA 2024-T3, the thermal strain was not sufficient to cause yield strength magnitude residual stresses due to its high yield strength and low elastic modulus [360]. It was necessary to more accurately measure or develop validated FE

172 models incorporating the softening effect in AA 2024-T3 welds in order to predict welding induced residual stresses and distortions as close as possible to the real values.

The softening effect was included in the FE models by considering the mechanical response of welds. Constitutive data for the various microstructural regions that make up the weld including the FZ and HAZ were determined via experimental measurements using the DIC in Chapter 4. First, the base metal material properties were used during the heating stage of the welding thermal cycle and then replaced by the weld constitutive behaviour during the cooling stage. No uncertainty existed regarding whether the data used were actually representative of the material in the weld:

(ii) Residual stresses in W47 were measured and calculated using neutron diffraction technique from a reactor based diffractometer E3 at HZB. Figure 93 shows the residual stress distributions obtained from the {222} and {311} reflections for aluminium compared to the results from numerical simulations. The choice of diffraction elastic constants E and ν depended on the crystallographic hkl plane used for strain measurements. Since crystallographic anisotropy of the elastic constants is small for aluminium, the values of E=70.2 GPa and ν=0.35 were used for both the Al {222} and {311} reflections to calculate stresses. The stresses were calculated using three different methods. For the first set of measurements, local stresses were calculated based on strain free lattice parameters obtained from a stress free reference comb with thin teeth from the weld region where macroscopic stresses are known to relax. Alternatively, a far field global reference value obtained from a position which was regarded as stress free was used to calculate stresses. Finally, plane stress condition was assumed, where the z stress component was set to zero and therefore, the x and y stresses were determined using the biaxial formula. Figure 93 shows that the stresses obtained from the {222} and {311} peaks were not quite the same. The cause of such difference was thought to be due to the influence of microscopic stress on the stress results. While the {311} diffraction peak gives good measure for the macro-stresses, the {222} peak on the other hand, may be affected by intergranular stresses which cause the elastic lattice strain for this particular reflection to differ from the true macroscopic elastic strain. Considerable differences were observed between the two reflections for the y stresses, whereas, the differences for the x and z stresses were relatively smaller. It was obvious from both results that the contribution of the x and z residual stresses were much smaller than that from the y residual stresses, within the range of around ±100 MPa.

The results obtained from plane stress condition, globally applied reference parameter and local stress free parameters were all in good agreement and matched the FE predictions well. There were no significant differences among these results and even the plane stress assumption gave reasonable stress distributions. The results from the {311} reflection

173 matched the FE results better than those from the {222} reflection, which over predicted the y stresses in the weld region.

ai) bi) 400 400 FEA FEA σ11 Ref σ σ11 Ref σx σ11 Global x σ11 Global 300 300 σ11 Plane stress σ11 Plane stress

200 200

100 100

0 0

Transverse(MPa) stress

Stress Stress (MPa) Stress Stress (MPa) -100 -100

-200 -200 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 DistanceDistance from from weld weld centre center (mm) (mm) DistanceDistance from from weld weld centre center (mm) (mm) aii) bii) 400 400 FEA FEA σ22 Ref σy σ22 Ref σy σ22 Global σ22 Global 300 300 σ22 Plane stress σ22 Plane stress

200 200

100 100

0 0

Longitudinal Longitudinal stress (MPa)

Stress Stress (MPa) Stress Stress (MPa) -100 -100

-200 -200 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 DistanceDistance from from weld weld centre center (mm) (mm) DistanceDistance from from weld weld centre center (mm) (mm) aiii) biii) 400 400 FEA FEA σ33 Ref σz σ33 Ref σz σ33 Global σ33 Global 300 300

200 200

100 100

0 0

Normalstress (MPa)

Stress Stress (MPa) Stress Stress (MPa)

-100 -100

174 The {222} y stresses matched the FE results the best within the first ±5 mm from the weld centre when stress free lattice parameters from the comb sample were used, but indicated greater compressive stresses further away than the FE predictions. Using the plane stress condition also produced similar results, whereas, globally applied reference parameters led to further over prediction of y stresses by almost 100 MPa more than the FE predictions. The {222} stresses also showed larger uncertainties than the {311} stresses so it was unavoidable to take into account these variations. On the other hand, the {311} y stresses were in good agreement with the predicted stresses for all three calculation methods.

Both the magnitude and distribution of y stresses in and around the weld were well matched and only small variability of stress measurements was observed. The peak y stresses were measured to be close to 200 MPa at around ±3 mm from the weld centre, similar to those of the FE model and then dropped to zero at a similar rate at around ±6 mm from the weld centre. The minimum tensile y stresses of around 50 MPa were measured at the weld centre when using either global or local lattice parameters, which were lower than the FE stresses, whereas those from plane stress condition were similar, approximately 100 MPa. This proves the importance of using local stress free lattice parameters in order to capture precisely the lower y residual stresses due to softening of the weld metal. In contrast, the y stresses obtained under plane stress condition matched the FE predictions better than the other two in the compressive region far from the weld, whereas, the opposite was true for x stresses.

Residual stresses in W99 and W102 were measured and calculated using neutron diffraction technique from a pulsed spallation source ENGIN-X at ISIS. As discussed previously in Section 5.3.2, the residual stresses calculated using this instrument were not specific to a particular peak but instead obtained by fitting an average lattice spacing to the whole diffraction pattern using the Rietveld (or Pawley-Rietveld) technique [357] as shown in Figure 95. The strain calculated using this method is known to give a value close to the macroscopic strain in both elastic and plastic region so the macroscopic bulk elastic constants were used to determine the stress [358].

As it can be seen from Figure 94 and Figure 95, large stress gradients were present within a very narrow region around the weld and almost stress free in the rest of the regions. Again, the measured x and z stresses were very small, typically much less than ±100 MPa and in good agreement with the predicted stresses for both W99 and W102 under all three conditions. Figure 95 shows that the specimen W102 clearly had some compressive stresses in the weld as much as -100 MPa, whereas, predicted stresses showed almost no compressive behaviour in this region. The y stresses derived from neutron diffraction measurements were also very similar to the numerically determined stresses. The peak tensile y direction residual stresses were in the range between 200 and 300 MPa depending on the method used to calculate the

175 stresses. A lower tensile y direction residual stress at the weld centre of around 60 MPa was obtained when using local stress free lattice parameters compared to over 100 MPa when using a global value or assuming plane stress condition. Similarly, higher peak tensile y residual stresses of around 250 MPa were obtained around the weld when using the local stress free lattice parameters compared to the other two cases which were less by around 50 MPa.

a) b) 400 400 σ11 FEA σ22 σx σ11 Ref σ33 σ11 Global 300 300 σ11 Plane stress

200 200

100 100 Stress (MPa)

0 0

Stress Stress (MPa)

Stress Stress (MPa) Transverse(MPa) stress

-100 -100

-200 -200 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 DistanceDistance from from weld weld centre center (mm) (mm) Distance from weld centre (mm) Distance from weld center (mm) c) d) 400 400 FEA FEA σy σ22 Ref σz σ33 Ref σ22 Global 300 300 σ33 Global σ22 Plane stress

200 200

100 100

0 0

Normal(MPa) stress

Stress Stress (MPa)

Stress Stress (MPa) Longitudinal Longitudinal stress (MPa)

-100 -100

-200 -200 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 DistanceDistance from from weld weld centre center (mm) (mm) Distance from weld centre (mm) Distance from weld center (mm)

W102 was welded with filler metal in order to control the weld microstructure and improve the weld quality. It was identified in Chapter 3 that filler metal has a considerable influence on the weld metal chemical composition so the effect of weld metal chemical compositional and microstructural variations on intergranular stresses must be accounted for if accurate strains are to be measured in this region [344]. Figure 95 shows that when such variations were not

176 considered in the stress calculations as in the case of using a globally applied reference parameters, the y residual stresses at the weld centre were no longer tensile in nature but instead close to zero. On the other hand, when local stress free lattice parameters from reference comb were used, then the y stresses become considerably higher, as high as 100 MPa. Likewise, under plane stress condition, stress magnitudes over 100 MPa were found at the weld centre. For this reason, it was crucial to take into account the chemical compositional and microstructural variations across the weld when welded with filler metal by choosing the right stress free reference lattice parameters which remove these effects.

a) b) 400 400 σ11 FEA σ11 Ref σ22 σx σ33 σ11 Global 300 300 σ11 Plane stress

200 200

100 100 Stress (MPa)

0 0

Transverse(MPa) stress

Stress Stress (MPa) Stress Stress (MPa)

-100 -100

-200 -200 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld center (mm) Distance from weld center (mm) c) d) 400 400 FEA FEA σ22 Ref σ33 Ref σ22 Global 300 σy 300 σz σ33 Global σ22 Plane stress

200 200

100 100

0 0

Normalstress (MPa)

Longitudinal Longitudinal stress(MPa)

Stress Stress (MPa) Stress Stress (MPa) -100 -100

-200 -200 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld center (mm) Distance from weld center (mm)

(iii) Figure 96 and Figure 98 show the residual stress distributions in fillet welded single and triple T-joint specimens measured using neutron diffraction technique from a reactor based diffractometer SALSA at ILL. The stresses were calculated using two different methods, one

177 using a far field global reference lattice parameter obtained from the edge of the specimen and the other assuming plane stress condition. Unfortunately, for these specimens, local stress free not reference lattice parameters in the weld were not available because a reference comb was prepared at the time of measurements. However, since the measurements were taken at mid-thickness of the sheets which were below the fillet weld regions, at least there was no microstructural or chemical compositional variation at these points.

a) b) 300 300

Measurement σ11 FEA Positions Fillet weld Stiffener σ22 250 ND 250 σ33 σx ND (Plane stress)

Z (Normal) 200 150 mm 200 300 mm 221 mm 150 Y (Longitudinal) X (Transverse) 150

100 100

50 50

Stress (MPa) 0 0 Stress Stress (MPa)

-50 Stress (MPa) -50 Transverse stress(MPa)

-100 -100

-150 -150

-200 -200 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 DistanceDistance from from specimenweld centercentre (mm) (mm) DistanceDistance from from specimen weld centre center (mm) (mm) c) d) 300 300 FEA FEA 250 ND 250 ND σy ND (Plane stress) σz 200 200

150 150

100 100

50 50

0 0

-50 Normal(MPa) stress -50

Stress Stress (MPa)

Longitudinal Longitudinal stress(MPa) Stress Stress (MPa)

-100 -100

-150 -150

Figure 96 shows that the peak x residual stresses in the single T-joint specimen determined from the global lattice parameter had almost 50 MPa larger stress magnitudes in the weld region both in tension and compression of around ±150 MPa compared to those in butt welded specimens which were on average less than ±100 MPa. On the other hand, the x stresses calculated under plane stress condition were much smaller in the range of only around ±50 MPa. The trend displayed by both measured and simulated stress distributions were very

178 similar but with the maximum difference of around 100 MPa in the peak x stresses close to the specimen centre when the global parameter was used. In the case of z stresses, they were equal to zero under plane stress condition, so only those determined using the global parameter is shown in Figure 96 were compared to the predicted z stresses. The maximum measured z stresses are approxmiately equal to -100 MPa in compression and slightly greater than 150 MPa in tension, which were in good agreement with the predicted stresses in compression but again around 100 MPa higher in tension. Still, as it can be seen from Figure 96 the x and z stresses were small relative to the material’s yield strength at room temperature.

The distribution and magnitude of y stresses determined using the global parameter matched the predicted stresses very well, as illustrated by the width of the large tensile stress gradient around the specimen centre within the first ±5 mm and the maximum y stress magnitude as high as 280 MPa near the weld, which was much higher than those measured in butt welded specimens and comparable to the yield strength of 345 MPa. On the other hand, the longtiduinal stresses calculated assuming plane stress conditions showed around 100 MPa lower peak tensile stresses and also -50 MPa more compressive stresses in the rest of the specimen than the other measured and predicted stresses. This meant that unlike butt welded speciments, the fillet welded T-joint specimens in the vicinity of the stiffener were not quite in the plane stress state.

Figure 97 shows that for all three stiffeners in the tripe T-joint specimen, the y residual stresses on the left side of each stiffener which was welded after the right side were around 20 MPa higher. It was thought that the heat conduction during the second weld pass from the left to the right side of the stiffness resulted in partially relieving and redistributing the residual stresses in the right region. It was found that the tensile region in the y direction reached approximately up to 10 mm across the specimen centre which quickly became compressive further away to around -50 MPa and then levelled off towards the edge of the sheets.

Residual stresses predicted by numerical simulation showed a similar distribution to those measured by ND. The FE model reasonably predicted the magnitude of the tensile residual stresses. However, the differences in the y stress magnitudes between the two sides of the stiffeners described above were not obvious from the predicted stresses. Although the measured and predicted stresses were in close agreement for the mid-stiffener, some differences were found with the other two side stiffeners. The peak y stresses in the left stiffener which was welded last were greater than those in the middle stiffener by almost 50 MPa. In contrast, those in the right stiffener welded second were lower than those in the middle and left stiffeners by more than 30-80 MPa. Such asymmetrical stress distributions along the weld centre line in this specimen indicated that welding sequence may significantly

179 influence the resultant residual stresses. As predicted by the FE model, the magnitude and distribution of compressive stresses measured using ND in the y direction were very small. This was possibly due to the fact that only small compressive stresses were needed over a much wider volume relative to the very narrow weld zone in tension, to achieve a force equilibrium.

a) b) 350 350 σ11 Measurement FEA 300 σ22 300 σx Positions Stiffener ND ND (Plane stress) σ33 Fillet weld 250 250 25 mm

200 200 Z (Normal) 400 mm 380 mm 150 150 Y (Longitudinal) X (Transverse)

100 100

50 50

Stress (MPa) Stress Stress (MPa)

0 Stress (MPa) 0 Transverse stress(MPa) -50 -50

-100 -100

-150 -150 -200 -150 -100 -50 0 50 100 150 200 -200 -150 -100 -50 0 50 100 150 200 DistanceDistance from from weld specimen center centre (mm) (mm) DistanceDistance from from specimenweld centrecenter (mm) (mm) c) d) 350 FEA 350 ND FEA 300 ND (Plane stress) 300 ND

250 250

200 200

150 150

100 100

50 50 Normal stress(MPa)

Stress Stress (MPa) 0 0

Stress Stress (MPa) Longitudinal Longitudinal stress (MPa) -50 -50

-100 -100

-150 -150 -200 -150 -100 -50 0 50 100 150 200 -200 -150 -100 -50 0 50 100 150 200 DistanceDistance from from specimenweld center centre (mm) (mm) DistanceDistance from from specimenweld center centre (mm) (mm) Figure 97 Residual stress distributions a) across the entire width and in b) x, c) y and d) z directions measured using neutron diffraction technique and numerically simulated for triple T-joint specimen in {311} reflection, assuming plane stress or globally applied far field reference value (AA 2024-T3, T-joint)

In addition, residual stresses were measured along the height of the stiffener, starting from the bottom of the base sheet to the top of the stiffener. Figure 98 shows that the predicted stresses in all three directions showed a dip up to around 2 mm and then reached a peak in the range between 2 and 4 mm from the bottom. The y stresses were in tension in the base plate and at the bottom of the stiffener with the maximum value of more than 250 MPa, and compressive in the stiffener with the minimum of around -100 MPa, which tended to zero at the top end of the stiffener. The measured tensile x and z stress were also considerably large in the base sheet with peak values over 150 MPa at 2 mm from the bottom and almost 100

180 MPa greater than the predicted stresses. Compressive x and z stresses higher up the stiffener were well captured by the numerical results. In conclusion, significant y residual stresses were measured along the height of the stiffener up to around 5 mm from the bottom as well as some x and z stresses in the same locations. All these stresses diminished at heights greater than 5 mm.

350 σ11 (FEA) 300 σ22 (FEA) 250 σ33 (FEA) σ11 (ND) 200 σ22 (ND) σ33 (ND) 150 100 50

0 Stress Stress (MPa) -50 -100 -150 0 10 20 Distance along stiffener height (mm)

Figure 98 Residual stress distributions along the height of the stiffener of fillet welded T-joint specimens in x, y and z directions measured using neutron diffraction technique compared to numerically simulated residual stress (AA 2024-T3, T-joint) 5.5 Results and Discussion on Ti-6Al-4V (i) Surface residual stress measurements were performed using a low energy X-ray diffraction (XRD) technique on the top surface of the butt welded Ti-6Al-4V sheet at a depth of around 10 μm from the surface. Figure 99 shows the bi-axial residual stresses, x and y to the welding direction, measured using the XRD and obtained from FE models either with or without phase transformations.

a) b) 100 800 XRD XRD FEA 80 FEA 700 FEA (εtr) FEA (εtr) 60 600 σx σy 40 500

20 400

0 300

-20 200

Stress Stress (MPa)

Transverse(MPa) stress Longitudinal Longitudinal stress(MPa) -40 100

-60 0

-80 -100 0 10 20 30 40 50 0 10 20 30 40 50 DistanceDistance fromfrom specimen weld centre centre (mm) (mm) DistanceDistance from from specimenweld centrecentre (mm) (mm)

181 It was assumed that the residual stress distributions are symmetrical about the weld centre line and therefore, only half of the sheet was measured. Both the measured and simulated x residual stresses were found to be within the maximum range of ±100 MPa, which when compared to the yield strength of 1100 MPa were very small.

The FE predictions were in good agreement with the experimental measurements. The predicted stress distributions showed large variations only up to around 4 mm away from the weld centre line and then levelled off close to zero. The magnitude of tensile and compressive stresses around the weld were predicted to be much greater in the model with phase transformations compared to the model without phase transformations which only showed around ±20 MPa variations at most. The measured stress distribution showed greater stress magnitudes than the model without phase transformations but smaller than the model with phase transformations within the weld. The experimental measurements showed slightly more compressive stresses away from the weld than the predicted stresses but the differences were small. It was therefore concluded that the stresses in the x direction were relatively small in this thin sheet.

The y stresses in the direction of the weld were found to be significant even very close to the surface according to both simulated and experimental results. The simulated and experimental y stresses had approximately the same magnitude and distribution, reaching around 70% of the room temperature yield strength in the weld. Peak tensile y stresses of around 750 MPa were predicted and observed in a very narrow weld region within the first 2 mm from the weld centre which rapidly dropped by more than half at 3 mm and then to below 0 MPa at 4 mm from the weld centre and became weakly compressive over 50 mm from the weld centre. It seemed too conservative to assume in this case that yield magnitude residual stresses occur in the weld. Peak residual stresses as high as the yield strength are likely to occur when the thermal contraction strain becomes larger than the yield strain. Often in steels, the thermal strain can reach more than the yield strain and result in yield magnitude residual stresses. However, in the case of Ti-6Al-4V, the thermal strain was not sufficient to cause yield strength magnitude residual stresses due to its high yield strength and low elastic modulus [360].

The predicted y stresses at the top of the sheet were the maximum at the FZ/HAZ interface and roughly 50 MPa lower at the weld centre. While the same trend was observed in both models, the model with phase transformations predicted slightly lower stress magnitude at the weld centre and higher at the FZ/HAZ interface than the model without phase transformations. On the other hand, the measured peak tensile y stresses were located at the weld centre and the opposite trend was observed where the stress was approximately 50 MPa lower at the FZ/HAZ interface. Still, overall the FE predictions were in good agreement with the experimental measurements.

182 (ii) Figure 100 shows the residual stress and strain measurements obtained using the neutron diffraction technique compared to the results from numerical simulations. It was uncertain at first which reflection would be the best to use for measurements or the texture that would be expected. The detector was place to measure several peaks simultaneously including the {103}, {112} and {201} peaks. As the {103} peak is recommended in the ISO/TTA 3:2001 standard [361], it was included on the detector. As it can be seen from Figure 100 strains from all three reflections were measured in the x direction. On the other hand, different intensities were produced for different peaks in the y and z directions. Measurement of the {103} y strains was found to be problematic due to very weak or the absence of the {103} reflection. Similarly, while the {103} reflection was present in the z direction, the {201} was very weak. The likely cause of such result could be crystallographic texture in the weld metal induced during the solidification process on cooling. Texture can influence the peak intensity of a given hkl diffraction peak measured in a particular direction due to preferred orientation of certain crystallographic planes along certain macroscopic direction. It causes the measured relative intensities of the peaks from favourable oriented crystallographic planes to be higher than from less favourably oriented ones, and therefore, prevents the observation of one reflection in all the directions for which strain data are required.

Standford and Bate [362] studied transformation texture associated with martensitic transformation in Ti-6Al-4V. From any single β crystallographic orientation, there are 12 possible α orientation that may form during transformation. If transformation proceeds in certain orientations or variants more than others, it means variant selection (transformation texture) occurs. They conducted EBSD texture measurements and found that within each prior β grain, although all 12 variants of αʹ were formed, the fractions of variants were not uniform. They values were different from those calculated using equal variant probability, meaning that a significant variant selection had occurred during martensitic transformation. Also α and β in the parent material during hot rolling developed deformation textures characteristics of their crystal structure where a strong texture with basal poles aligned with the rolling direction typical of industrially rolled titanium alloys was observed.

Another problem when measuring residual stresses in Ti-6Al-4V was caused by the addition of aluminium and vanadium which reduced the average coherent scattering length of Ti-6Al- 4V but also increased its incoherent cross-section. As a result, the reflections even in texture free randomly oriented polycrystalline Ti-6Al-4V were weak and only few times larger than the background, making measurements difficult in this material [363]. For these reasons, it was necessary to use more than one reflection to calculate strains. Figure 100 shows that there was some difference between the {103}, {112} and {201} x strains. The {112} and {201} peaks agreed well with each other whereas, the {103} peak appeared to have larger values in the

183 weld which could be due to intergranular strain or different moduli values. Highly textured nature of Ti-6Al-4V resulted in lower intensity counts observed during neutron diffraction measurements for certain reflections where one less diffraction peak was observed in the radial and z directions than the x direction. Only the {112} and {201} peaks were detected in the y direction and there was a difference between the two y strains. It was found that the same difference of approximately 1000 micro-strain was also detected in the reference comb, meaning that the difference may also be due to intergranular strain.

The {103} and {112} peaks were detected in the z direction which appeared to have a systematic shift of the strain distribution with respect to each other. The {112} reflection was observed in all three directions and had the smallest value in the base metal for all directions, whereas, it seemed the {103} peak needed directional reference values instead of a global value for all strains.

As it was not possible to measure the z strain for the {201} reflection and the y strain for the {103} reflection due to texture, the y strain data from the {201} reflection was used to calculate stress for the {103} reflection and likewise, the z strain data from the {103} reflection was used to calculate stress for the {201} reflection. For the stress calculation, the y {201} peak and the z {103} peaks were chosen above the corresponding {112} peaks because of the possibility of intergranular strains in the {112} reflection. The experimentally determined stresses in the x and z directions were very close to zero for all reflections and the z stress being zero was what would be expected from a thin sheet. Also, the stresses in both directions agreed well with each other, suggesting a uniaxial stress distribution. In terms of the y stress in the weld, all three reflections showed the same stress distribution but slightly different peak tensile stress magnitudes. The highest stress magnitude was given by the {201} reflection of around 700 MPa, followed by the {103} reflection of around 650 MPa and the lowest for the {112} reflection of around 600 MPa.

184 a) b) 1000 7000 ε22 (201) 500 6000 ε22 (112)

0 5000

-500 4000

)

- - -1000 3000

-1500 2000

Strain Strain (x10 Strain Strain (x10 -2000 1000

-2500 0 ε11 (201) -3000 ε11 (112) -1000 ε11 (103) -3500 -2000 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre

c) d) 1000 800 112 σ11 (FEA) 700 σ22 (FEA) 500 σ33 (FEA) σ11 (ND) 600 0 σ22 (ND) σ33 (ND) 500

-500

) 6

- 400 -1000 300 -1500

200

Strain Strain (x10 Stress (MPa) -2000 100 -2500 0

-3000 ε33 (112) -100 ε33 (103) -3500 -200 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

e) f) 800 800 201 σ11 (FEA) 103 σ11 (FEA) 700 σ22 (FEA) 700 σ22 (FEA) σ33 (FEA) σ33 (FEA) σ11 (ND) 600 600 σ11 (ND) σ22 (ND) σ22 (ND) σ33 (ND) σ33 (ND) 500 500

400 400

300 300

200 200

Stress(MPa) Stress(MPa)

100 100

0 0

-100 -100

-200 -200 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

185 It was found that some unrelieved macro residual stresses retained in the y direction of the reference comb specimen so it was not completely stress free as shown in Figure 101. However, as the measured sheet was welded autogenously, the effect of compositional variation was small but microstructural variation across the weld could cause intergranular stress. It would therefore be necessary to measure again a stress free reference sample with the weld region in the coupon or correct the reference lattice spacing data by sin2ψ technique [364]. The FE predictions of all three principal stresses were in very good agreement with the residual stresses measured from the {201} reflection, whereas, the measured y stresses for the {103} and {112} peaks in and around the weld were less than the simulated values by around 100-150 MPa at most because the calculated {103} y stress was not the true value for the {103} reflection and also the {112} reflection was prone to intergranular stresses. Therefore, it was concluded that the {201} stresses which were in good agreement with the simulation results were the most reliable and proved that the FE predictions were accurate.

3500 3000 2500

2000 )

6 1500 - 1000 500

Strain (x10 Strain 0 -500 -1000 LONG 'REF' 201 -1500 LONG 'REF' 112 -2000 -15 -10 -5 0 5 10 15 Distance from weld centre (mm)

Figure 101 strain variation in the x direction in the reference d0 comb sample in the {112} and {201} reflections (Ti-6Al-4V, butt joint) 5.6 Conclusions The observed deviations in the peak y stress levels in Ti-6Al-4V and AA 2024-T3 specimens were mainly caused by the stress free lattice parameters used when processing the elastic strains from the neutron diffraction measurements. In particular, the magnitude of experimentally determined reduced y tensile residual stresses in the softened weld region of AA 2024-T3 specimens were largely controlled by the chosen stress free state. The use of a global far field lattice parameter resulted in under predicting these stresses for the butt weld AA 2024-T3 specimen welded with filler metal because the influence of chemical compositional and microstructural variations in the weld were not accounted for in this case, whereas, the use of local values or plane stress condition led to prediction of higher stresses.

186 The measured residual stresses were also dependent on the crystallographic hkl plane from which they were obtained from and the cause of such difference was found to be due to the presence of microscopic stresses. In the case of AA 2024-T3 specimens, both the {222} and {311} peaks were detected at the same time for W47, whereas, only the {311} peak was visible for the T-joint specimens. The {311} peak on average produced higher stress magnitudes due to the reason above. On the other hand, for the specimens W99 and W102, the entire diffraction spectrum was fitted with Rietveld refinements instead of conducting a single peak analysis so it was possible to use the macroscopic elastic constants and the results were in good agreement with the numerical results.

In the case of Ti-6Al-4V, the reflections were weak and only few times larger than the background due its highly incoherent cross-section, thus making neutron diffraction measurements difficult. In addition, texture in the Ti-6Al-4V weld also contributed to lower intensity counts observed during the measurements. As a result, only certain peaks were detected in certain orientations and therefore, different stress values were obtained for different reflections.

187 6 NUMERICAL SIMULATION OF WELDING RESIDUAL STRESSES AND DISTORTIONS IN AA 2024-T3 AND Ti-6Al-4V

6.1 Introduction Experimental measurement of welding temperature fields, residual stresses and distortions is time consuming, expensive and there may also be uncertainties associated with the measurements [365]. As a result, computational welding simulation is often used to predict the distribution and magnitude of transient residual stress fields and distortions in different types of materials and joints during and after welding. The advantage of welding simulation is that calculations are cheaper and faster than conducting experiments. It allows to retrieve quantities which are difficult to measure experimentally, and investigate systematically the influence of various aspects of welding such as welding parameters, restraints and subsequent processing by simulating ideal conditions which eliminate the practical limitations of conducting experimental studies for process optimisation [366]. However, computational simulation still requires experimental calibration on measurable parameters such as the weld pool geometry and temperature field in order to determine net heat input. It is important to validate the model experimentally before using it to conduct parametric studies. Therefore, model aided experiments increase the reliability of the results and can then be used for welding process optimisation and evaluation of in-service structural integrity and failure modes of welded structures due to welding residual stresses and in-service loads with confidence.

Figure 102 Schematic of the complex thermo-mechanical-microstructural phase transformation coupling interaction (figure taken from [367])

Welding is a process which involves many complex physical phenomena involving multi field interactions among metallurgical, thermal and mechanical changes of the weldment during welding as shown in Figure 102. Some of which include thermal stress induced by the welding thermal cycle and the accompanied melting, solidification or solid state phase transformation; transformation stress caused by local dilatation; heat generation due to inelastic deformation;

188 phase transformation accelerated by stress or strain; and the latent heat due to phase transformation [367].

Due to the limitation of available computational performance in the past, analytical approaches were often adopted. Ueda et al. [368] initially developed a simplified experimental and analytical combined method referred to as the inherent strain method, to predict the welding residual stresses by utilising the characteristics of the inherent strain distribution in a welded joint. Hill and Nelson [369] further developed this method. The method assumes that the effect of complex physical changes during welding on residual stresses and distortions is relatively small compared to the local plastic strains generated during cooling. Hence, the temperature and corresponding strain history during the early stages of welding are neglected [368]. The inherent strain model also assumes an axis-symmetric condition, meaning that it is not possible to predict the transient residual stress distributions near the weld start and end locations [370].

Increased computational powers and development of advanced numerical simulation techniques led to replacement of analytical solutions with finite element based numerical simulation techniques for weld modelling. The numerical approach is reaching its maturity level and frequently applied to solve industrial problems. Considerable amount of research on prediction of welding residual stresses and distortions has been performed through numerical simulations using the FEA method. A comprehensive literature review on the development of welding simulations before 2001 was done by Lindgren in three parts [371–373]. Similarly, the contributions up to 2004 were discussed by Yaghi and Becker [374] including a critical review on history of welding simulation, discussion on published literature, modelling considerations such as the effects of different parameters, material properties, geometric non-linearity and meshing techniques; and proposed recommendations for future work. Another literature review on development of FEA for laser welding up to 2012 was done by He [375] in terms of process, damage modelling, fatigue behaviour, dynamic characteristics and laser hybrid welding. Numerical issues such as materials modelling, meshing procedure and failure criteria were also discussed with case studies.

Prediction of welding residual stresses and distortions is still a complex and difficult task as mentioned above because it involves modelling a complex non-linear elastic-plastic analysis. It is also required to account for a moving heat source; temperature dependent material properties; plastic flow; volumetric expansion during metallurgical transformations; filler metal deposition; and thermal and mechanical boundary conditions during welding. This means that welding simulation can be very computational resource heavy and demand large data storage. As a result, it was inevitable in the past to make several assumptions, approximations and simplifications while ensuring high enough accuracy of the predictions.

189 For accurate predictions, high quality temperature field and material constitutive data and appropriate boundary conditions are required. It is also important to include the effect microstructure evolution due to phase transformations changes occurring in the weld metal and the heat affected zone in the models for Ti-6Al-4V and softening in AA 2024-T3, which may lead to changes in the peak residual stress values. Zhu and Chao [376] studied the effects of temperature-dependent material properties on transient temperature, residual stress and distortion in finite element simulation of a welding process. It was determined that the yield stress is the key mechanical property so the influence of temperature on the yield stress must be considered in order to obtain correct results. On the other hand, thermal conductivity, thermal expansion coefficient and elastic modulus have relatively smaller effect on the distribution of the residual stress and distortion after welding.

When temperature dependent material properties are unavailable, using room temperature material properties except for the yield stress produced reasonable predictions of the transient temperature fields, residual stress and distortion. They stated that the results obtained by using room temperature elastic modulus are significantly more accurate than those obtained using its average value over the welding temperature history [376]. According to Deng and Kiyoshima [377] the final residual stresses at the weld and its vicinity are not affected by the initial residual stresses and only determined by the welding process, whereas, at greater distance from the weld centre, the influence of the initial residual stresses on the final residual stresses becomes larger. Apostol et al. [378] investigated the influence of input welding parameters on residual stresses and distortions using FEA and concluded that laser power, welding speed and spot diameter have a considerable influence on their magnitude. It was found that when the laser beam power increases, displacements and residual stress values also increase. In contrast, they both decrease as the welding speed and spot diameter increase.

In most cases when calculating macroscopic residual stresses and distortions in welded structures, fluid flow phenomenon is not explicitly modelled as the heat input into the weld and the weld pool geometries have negligible effects on the deformation and stress field, so the model does not predict weld pool geometries but instead, uses the weld pool geometries determined from experiments as input parameters for calibration. The effect of the fluid flow is only significant when geometrical changes close to the weld pool are of primary interest [333].

Often when simulating thin sheet welded structures where the temperature distribution is relatively constant through thickness, a plane stress assumption is used which implies that temperature and deformation are constant through thickness. However, with this assumptions, all stresses are in-plane so out-of-plane angular distortions cannot be predicted. In addition,

190 boundary conditions on the plane surface such as thermal heat loss due to convection and radiation cannot be assigned in this case.

Early simulations were based on two dimensional (2D) cross-section analysis, which predicts residual stresses according to plane strain assumption, corresponding to rigid end constraints but is unable to determine out-of-plane deformations [379]. In addition, for materials exhibiting phase transformations, a 2D model predictions show significant differences with experimental measurements due to the plane strain condition in which plane sections remain planar and out-of-plane normal and shear strains are neglected. Still, the 2D models can be useful in situations where in-plane stresses are of main interest [380].

The 2D FE model generally predicts a higher peak temperature in the weld and a lower temperature in the areas further away than that obtained from the 3D FE model because the heat transfer in the welding direction is not considered [381]. Since welding is a transient process, cooling does not occur uniformly at all positions along the weld so the change in stiffness as a result of the welding thermal history cannot be realised in the 2D model. Michaleris et al. [382] claimed that in a 2D model, full penetration weld can become unstable due to the decrease of stiffness at high temperatures, whereas, a 3D model can accurately predict the welding distortion as the welding heat source travels along the weld path.

Due to the nature of the welding process, a 2D model is not adequate in many practical problems so a full scale 3D model consisting of finite solid elements with time incremental representation of the moving heat source is necessary [379]. According Oddy et al. [383], prediction of welding residual stresses and distortions greatly rely on accurate prediction of welding temperature field and such prediction requires a nonlinear, transient 3-D analysis. The 3D model is not limited by axisymmetric states, plane stress or plane strain states and material behaviour discontinuities as it is the case in the 2D model. It is also possible to predict out-of- plane deformations in all geometrical directions like angular distortion and cambering that cannot be obtained from 2D analysis.

In a fully coupled analysis, both thermal and mechanical behaviours are analysed simultaneously incorporating the effects of the latent heat of microstructural transformation and the heat produced by mechanical work due to the dissipation energy of plastic deformation. Several investigators namely Lindgren [372] and Dong [384] showed that it is admissible in predicting welding residual stresses and distortions with minimal loss in accuracy by uncoupling the welding simulation into sequential thermal transient and elastic-plastic mechanical analyses, The predicted results do not differ much from those given by a coupled analysis. Oddy et al. [383] stated that the heat generated by plastic deformation is significantly smaller than the welding heat. According to Pietzsch [385] this influence of mechanical work

191 is about four orders of magnitude lower than the welding heat input so it is a common practice to neglect such effect in welding simulations. In an uncoupled analysis, the thermal history is predicted first by the thermal analysis independent of the residual stresses and microstructural states, which is then transferred to the subsequent mechanical analysis as input while introducing temperature dependent material properties [334]. The benefit of using the uncoupled analysis is reduction in computational time and possibility to verify the thermal analysis independently prior to thermo-mechanical analysis. Although, microstructural changes in most cases are neglected in simplified models, the microstructural transformation in Ti-6Al-4V was taken into account in this investigation by considering the material properties variations with respect to the welding temperature field.

6.2 Materials and Welding Procedures The materials considered in this investigation were thin sheets of 3.2 mm thick AA 2024-T3 and 2.0 mm thick Ti-6Al-4V, the details of their material properties and chemical compositions are given in Section 3.1.1. Single pass welding of these two materials was performed using fibre laser in keyhole mode with the identical laser beam parameters, welding machines and clamping conditions, as given in Section 3.1.2 to produce scaled down specimens of dimensions and geometry as shown in Figure 88 of Section 5.3.2, which were large enough to generate the required results regarding temperature, residual stress and distortion.

Additional clamping bars were introduced when fillet welding T-joints in order to hold the stiffeners in position during welding. The welding parameters used are listed in Table 13. According to Table 13, AA 2024-T3 was welded in three different weld joint configurations which included butt welds welded with or without filler metal and fillet welded T-joints with the addition of filler wire on both sides of the fillet. AA 4043 filler metal of 1.0 mm in diameter was used to prevent hot cracking in AA 2024-T3. For T-joint welding, the laser beam was tilted by 30° (seam angle) from the horizontal plane and focused on (no defocus) the surface of the workpiece and argon gas shielding was used at a flow rate of 20 l/min. Ti-6Al-4V was only butt welded using a single set of optimised welding parameters as identified in Section 3.

Table 13 Welding parameters used to perform fibre laser welding experiments on butt welded and T-joint fillet welded sheets

Laser Welding Focal Wire feed Thickness Material Weld type power speed position rate (mm) (kW) (m/min) (mm) (m/min) AA 2024-T3 3.2 Butt 4.9 3.0 +4 - AA 2024-T3 3.2 Butt 2.9 1.5 +4 - AA 2024-T3 3.2 Butt 4.9 3.0 +4 5.2 AA 2024-T3 3.2 Fillet 3.5 1.8 0 5.0 Ti-6Al-4V 2.0 Butt 2.1 2.1 +4 -

192 The temperature dependent thermo-physical and thermo-mechanical material properties of AA 2024-T3 are presented in Figure 103 and Figure 104, and those of Ti-6Al-4V are presented in Figure 105 and Figure 106. An annealing temperature of 413°C, solidus temperature of 502°C and liquidus temperature of 638°C was used. The information was required for the use in numerical simulation models. The temperature dependent tensile properties of AA 2024-T3 were determined by uniaxial tensile testing standard specimens at temperatures ranging from 20°C to 500°C, in a homogenised condition, using the Gleeble 3800 thermo-mechanical simulator. The ratio of the gauge length (25 mm) to the gauge width (5 mm) was 5:1, complying with the standards. The specimen was clamped within copper grips and a pair of thermocouples was welded to each specimen to monitor the temperature of the specimen within the gauge length and a C-gauge transducer was used to measure the change in width of the specimen at its centre which enabled calculation of the strain. A constant strain rate of 0.001 s-1 was used. Literature values for the temperature dependent tensile properties of AA 4043 were used as shown in Figure 104 [386].

80 3.0 1400 160

70 140 2.5 1200

60 120 1000

2.0 ) 50 3 100 800 E 40 ν 1.5 c 80 α ρ 600 k

30 Poisson's ratio 60

1.0 Density (g/cm Specific heat (J/kgK) Elastic Elastic modulus(GPa) 400

20 40 Thermal conductivity Thermalconductivity (W/mK) 0.5 10 200 20

0 0.0 0 0 0 100 200 300 400 500 600 0 100 200 300 400 500 600 Temperature ( C) Temperature ( C) Figure 103 Temperture dependent thermo-physical material properties of AA 2024-T3 used in the finite element model

500 100 20°C 27 °C 100°C 100 °C 80 400 200°C 200 °C 300°C 300 °C 400°C 300 60 500°C

200 Stress(MPa) Stress(MPa)

100 20

0 0 0.0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 Strain Strain Figure 104 Temperture dependent thermo-mechanical material properties of a) AA 2024-T3 and b) AA 4043 used in the finite element model (figure taken from [386])

193 Literature values were used for the temperature dependent thermo-physical material properties of Ti-6Al-4V as shown in Figure 105. The elastic modulus, Poisson’s ratio and density values were obtained from JMatPro simulation software, while, specific heat and thermal conductivity values were obtained from Boivineau et al. [387] and Mills et al. [388].

The solidus temperature of Ti-6Al-4V (Tsolidus) was chosen to be equal to 1878 K, β transus temperature (Tβ) equal to 1248 K, liquidus temperature (Tliquidus) equal to 1928 K and boiling point (Tb) equal to 3315 K according to the values given in Elmer et al. [281] and Rai et al.

[389]. As it can be seen from Figure 105, sudden changes in these properties occur at Tβ, and

Tliquidus. The latent heat effect associated with the release or absorption of energy upon solidification or melting as a result of phase transformation was considered and so the latent heat of fusion (Δhfus) was chosen to be equal to 286 kJ/kg and the latent heat of evaporation

(Δhevap) equal to 9830 kJ/kg [281,389].

The temperature dependent tensile properties of Ti-6Al-4V were determined using the same method as mentioned above for AA 2024-T3, as shown in Figure 106. A wider range of temperature was used from 20°C to 1000°C and two different strain rates of 0.001 s-1 and 0.01 s-1 were used at higher temperatures to investigate the effect of varying the strain rate on the mechanical behaviour of Ti-6Al-4V. The tensile testing was conducted under a 10−3 torr vacuum condition to minimise the oxidation of titanium alloys at higher temperatures. a) b) 140 5.0 900 60

4.5 800 120 50 4.0 700 100 3.5

600 40

) 3 3.0 80 500 2.5 30 60 400

2.0 Poisson's Poisson's ratio

Density Density (g/cm 300 20 Specific heat (J/kgK) Elastic Elastic modulus(GPa) 40 E 1.5

200 Thermalconductivity (W/mK) ν 1.0 10 20 ρ c 0.5 100 k 0 0.0 0 0 0 500 1000 1500 2000 0 500 1000 1500 2000 2500 Temperature ( C) Temperature ( C)

Figure 105 Temperture dependent thermo-physical material properties of Ti-6Al-4V used in the finite element model [387,388]

194 1200 20°C, 0.001/s 200°C, 0.001/s 400°C, 0.001/s 1000 400°C, 0.01/s 600°C, 0.001/s 600°C, 0.01/s 800°C, 0.001/s 800 800°C, 0.01/s 1000°C, 0.001/s 1000°C, 0.01/s

600 Stress(MPa) 400

200

0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Strain

Figure 106 Temperture dependent thermo-mechanical material properties of Ti-6Al-4V used in the finite element model

Temperature field measurements in the vicinity of welded joints during and after welding were performed at different positions from the weld centreline as shown in Figure 107using K-type thermocouples (Chromel/Alumel) with a diameter of 0.5 mm, an accuracy in temperature measurement of around 1 C, sensitivity of 41 μV/ C and response time of 20 ms for temperatures up to 1350°C [390], in order to obtain sufficient information for the calibration of the heat source and thermal boundary conditions in numerical simulation models. Several thermocouples were spot welded each at a single point on selected positions, to minimise the influence of significant thermal gradient involved due to directional heating of the workpiece during laser welding, and then connected to multi-channels data logging system to record the transient welding temperature history with a measurement frequency of 200 Hz. It was found to be difficult to precisely monitor the temperature history within the FZ or on the weld centreline due to insufficient maximum operating temperature of the thermocouples used, so instead measurements were taken adjacent to the weld centreline.

Figure 107 Positions of thermocouples during welding

195 6.3 Finite Element Model Geometry and Mesh Table 14 shows the details about the mesh for different welded components including the number of nodes and elements, element type used for thermal and mechanical analyses.

Table 14 Mesh details for butt welded and T-joint fillet welded plates

Thermal Mechanical Nodes Elements Type Nodes Elements Type Butt 527693 115800 DC3D20 137410 115800 C3D8R 1 Stiffener 466936 104524 DC3D20 117625 104524 C3D8R 3 Stiffener 1328550 293088 DC3D20 335900 293088 C3D8R

In the thermal analysis, 2nd order 3D 20 node quadratic diffusive heat transfer brick hexahedron elements (DC3D20) were employed. In the mechanical analysis, 1st order 3D 8 node linear hexahedron reduced integration elements (C3D8R) were used. The finite element shape function of the displacements in the mechanical analysis should normally be one degree higher than that of the thermal analysis. Reduced integration elements, C3D8R use a lower order integration to form the element stiffness and provide better approximation of the real behaviour. It is preferable to use reduced integration instead of full integration for non-linear displacement based problems in order to avoid over-estimating the stiffness matrix and getting singular behaviour due to plastic deformation and incompressibility (shear locking). The integration points of the C3D8R elements are located at the centre of the element so small elements are required to resolve the steep temperature and stress gradient at boundaries. With these elements, hourglass effect can occur which causes the elements to severely deform arbitrarily as a result of the zero energy degrees of freedom, while the stress field is unchanged since they do not lead to any stresses. Normally, hourglass control is automatically activated for this element type where artificial stiffness is applied to restrict these stress free zero energy nodal displacements.

In order to transfer the nodal temperature data from the thermal results file that was generated during the thermal analysis to mechanical model for the mechanical analysis, the same finite element mesh was used as shown in Figure 108 with the same number of elements but different element types. Convergence studies on mesh density were done prior to this investigation [391] to ensure that a sufficiently refined mesh is used to yield accurate simulation results and to examine the solution’s sensitivity to element type and boundary conditions, the details of which are shown in Table 14 and Figure 108, with the smallest element size of 0.5 x 0.5 x 0.5 mm3.

196 a)

Figure 108 Mesh details of a) butt welded plates, b) T-joint fillet welded plates with one stiffener and c) and three stiffeneres with a user controlled refined mesh

197 The mesh converged when further mesh refinement produced a negligible change in the solution. It was found that a model with 115,800 elements for butt welds, a model with 104,524 elements for one T-joint fillet welds and a model with 293,088 elements is more than sufficient to accurately represent stress values in the weld region. A very fine mesh density was required in the weld region due to the small diameter of the moving heat source which traverses along the weld at specific time steps and the resulting high temperature and stress gradients in the fusion zone and the heat affected zones. The element size increased progressively with increasing distance from the weld centreline. Figure 108 shows mesh refinement for two different element sizes, either 1-2 or 3-5 mesh transition. At least three to four stages of mesh refinement were used by partitioning and transforming the mesh to create a user controlled refined mesh so that the computation time can be minimised and still be able to achieve accurate results. Even with such mesh refinement, thermal analysis of butt welds took on average over 2 hours, 2-3 hours for one T-joint fillet welds and 40 hours for three T-joint fillet welds per run; and mechanical analysis took 15 hours, 17-25 hours and over 170 hours per run, respectively, using 20 Intel Xeon CPUs in the Imperial College High Performance Computing Service, Cx1 cluster.

6.4 Thermal Analysis Analytical solution for simulating the transient temperature distribution in arc welding from a moving heat source was initially developed by Rosenthal in the 1940s for a point, and 2D and 3D line source in an infinite solid [392] and further developed by Rykalin [393] in the 1950s. They used the work of Carslaw and Jaeger about heat conduction in solids [394] and applied it to welding problems. Several assumptions and simplifications were made in the analytical solutions in order to solve the partial differential equation of heat conduction. It was assumed that the material properties are constant such that the material is isotropic and homogenous, no phase transformations occur and remain as solid regardless of time and temperature. In addition, the thermal properties including the thermal conductivity, density and specific heat capacity were assumed to be temperature independent, meaning that the results depend on the temperature at which these properties were obtained.

The welding process was assumed to be in steady state and adiabatic boundary conditions were assumed for all surfaces. The moving heat source was considered to be a zero volume point or line source on an infinite plate as mentioned above [395]. The problem with using Rosenthal’s models for the representation of the heat source for laser beam welding is that even though a reasonable approximation of the temperature in the far field can be made, they are prone to serious errors for predicting the temperature field in or near the weld. In these models, the temperature at the heat source is assumed to be infinite so as the heat source is approached the error increases. Later on Rybicki et al. [396] developed multiple point heat

198 source models which gave better approximation of the transient temperature distribution, and Pavelic et al. [397] was the first to develop a disc model with Gaussian distributed surface heat flux. However, the surface disc model is only effective for low power density welds with limited penetration or thin plates and not suitable for full penetration laser beam welds. Goldak et al. [398] developed a distributed power density double ellipsoidal heat source model to overcome these issues so that deep penetration welds with complex weld pool geometries can be modelled.

Figure 109 3D conical Gaussian type heat source model used for both butt welds and T-joint fillet welds

Figure 109 shows a 3D conical model which has a Gaussian power density distribution radially and a linear distribution axially produces more accurate results for deep penetration laser beam welds [398], and therefore, this heat source model was used in the numerical model for the thermal analysis. The conical heat source has a maximum heat flux at the top surface and a minimum on the bottom surface as expressed in Equation 18.

2  r    3   r0  Equation (18) qr (r, z)  q0e where r0 is the radius of the heat source at a specific height z, q0 is the maximum volumetric power density and r is the current radius of the interior point, which according Equation 19 is:

1 2 2 2 Equation (19) r  (x  x0 )  (y  y0  vt) 

The distribution parameter, r0, decreases linearly from the top to the bottom of the conic region as shown in Equation 20.

(ze  z) r0 (z)  re  (re  ri ) Equation (20) ze  zi

199 where ze is the top surface and zi is the bottom surface of the cone region. The maximum volumetric power density can be obtained by integrating the volumetric heat flux, Q over the body and rearranging it to the total heat input as shown in Equation 21.

Q  q ( x, y,z )dxdydz  r U  q ( r, )rdrddh  r U 3r 2 H 2 r0  Equation (21) 2  q e r0 rdrddh    0 0 0 0 9Qe3 q0  3 2 2 ( e  1)( ze  zi )( re  reri  ri ) where H=ze-zi and h=z-zi, e is the base of the natural logarithm and η is the heat source efficiency. Heat source efficiency or energy transfer efficiency is defined as the ratio of the energy absorbed by the irradiated materials to the laser power output. The efficiency is equivalent to the absorptivity of the metal during conduction mode welding. On the other hand, in the keyhole mode welding regime, the energy absorption during laser welding of various materials is not affected much by differences in composition and physical properties, so it can be said that it is independent of material [399]. The energy transfer efficiency used in the thermal model simulation was in the range of 75-80% after calibration with experimental thermocouple measurement of welding transient temperatures. The value was established by matching the thermal analysis results to experimental measurement of welding thermal histories. Several iterations were made to fit the heating and cooling rate, start and end temperatures and the maximum temperature reached. The process first involved matching the simulated weld transverse cross-section profile to experimental macrographs by changing the thermal efficiency value and heat source distribution parameters. Secondly, the welding thermal cycle was calibrated by changing the values for the ambient temperature; and conductive, radiative and convective heat loss using the *RADIATION and *FILM options in ABAQUS. The cooling rate was largely sensitive to the heat conduction parameters.

In order to simulate the moving heat source for T-joint fillet welding at an angle, a local coordinate system (x, y, z) was defined by translation and rotation transformation at the starting point of welding as shown in Equation 22.

푥 푐표푠 45 푠𝑖푛 45 0 푋 − 푋0 [푦] = [− 푠𝑖푛 45 푐표푠 45 0] [푌 − 푌0 ] Equation (22) 푧 0 0 1 푍 − 푍0 where X0, Y0 and Z0 are the coordinate values of a point in the global coordinate system (X, Y, Z) as shown in Figure 109.

200 Prediction of the weld pool size and shape requires calculation of the weld pool dynamics by solving full solution of Navier-Stokes equations and energy equations. However, as mentioned previously, due to the complexity of the physics of the welding process, a different approach was taken to simplify the solution based solely on the heat equation, where the heat source was simulated instead by prescribing volumetric heat flux input to the weld model. The heat equation derived from Fourier’s law and conservation of energy can be expressed as shown in Equation 23 and is used to solve the transient temperature field (T) in time (t) and space (x, y, z):

휕푇 휕 휕푇 휕 휕푇 휕 휕푇 ρ퐶 = (푘 ) + (푘 ) + (푘 ) + 푄̇ E Equation (23) 푝 휕푡 휕푥 휕푥 휕푦 휕푦 휕푧 휕푧 푉 where T is the temperature, ρ is the density, 퐶푝 is the specific heat, k is the thermal conductivity

3 and 푄̇푉 is the volumetric heat flux in W/m . In order to solve the heat equation, the thermal conductivity, density and specific heat must be specified. For a material which undergoes phase transformations, latent heats of phase transformation and microstructural evolution must also be considered [247].

The laser beam welding process was simulated in the thermal analysis with the 3D conical Gaussian heat source model, moving in the welding direction at a specified constant welding speed. The thermal analysis consisted of two main stages. The first stage involved applying a volumetric heat flux to realise the welding process and in the second stage, the heat source was removed to let the workpiece cool down due to set thermal boundary conditions to steady state conditions. Representation of the heat source with complex geometries was done using the ABAQUS user subroutine DFLUX programmed in FORTRAN. The DFLUX subroutine calculated the position of the heat source with respect to time, welding speed and nodal coordinates, and the volumetric heat flux at each integration point as a function of the heat source power, and the radius and depth of the affected area [400]. Gery et al. [381] investigated the influence of welding parameters on temperature distributions during welding and found that welding speed, heat input and heat source parameters have important effects on the weld geometries, and peak temperatures in the weld, which consequently affect the transient temperature distributions. Variations in the heat source parameters displayed a non- linear effect on the maximum temperatures reached in the weld and temperature distributions in the adjacent regions. Therefore, it is important to model the heat source accurately in order to predict welding residual stresses and distortions correctly. During laser welding, the maximum weld pool temperature can reach approximately the boiling point of the welded material as reported by Khan and Debroy [401], a value which was close to its boiling point of around 3000°C for stainless steel. Kroos et al. [39] measured the temperature of the keyhole wall and concluded that the temperature exceeded the boiling point by around 100 K.

201 Calibration of the heat source such as the peak temperature, shape and dimensions of the weld pool was performed by adjusting welding process parameters and heat source parameters for each weld FE model to fit the experimentally observed macrograph of a weld cross-section showing the FZ and the HAZ boundaries through thickness [376,402,403]. The calculated temperature fields after matching the macrograph were then evaluated against thermocouple measurements of the transient temperature fields captured during real welding experiments at corresponding locations for further calibrations. Thermocouples were directly mounted on the surface of the workpiece at various locations across weld path. Around four to five locations transverse to the weld were recorded using K-type thermocouples as described in Section 6.2. It was difficult to place the thermocouple very close to the weld without exceeding temperatures above the limit of the thermocouple so the temperature histories were measured a small distance away from the FZ, in the HAZ and a few points further away. The disadvantage of using thermocouple is its positioning, response time and thermal capacity which may all affect the measurement, especially near the weld where very high temperature gradients were present. An alternative measurement method is to use infrared thermography, an optical measurement technique which does not have the issues of thermocouples and has the advantage of measuring the temperature on whole field basis but the high equipment cost and calibration issues restrict its use.

Initial and ambient temperatures of the FE model at the beginning for all simulations were set to 20°C. Thermal boundary conditions were set to model heat transfer due to convection in air, radiation from the surface of the workpiece to ambient air according to the Stefan-Boltzman

202 relation and conduction from the workpiece to the mild steel support as shown in Figure 110 for butt welded and T-joint fillet welded plates. Heat transfer coefficients were adjusted in order to calculate the similar temperature fields as those measured experimentally using thermocouples.

Equation 24 and Equation 25 define heat loss due to surface convection and radiation, respectively, as boundary conditions.

푞푐표푛푣 = ℎ푐표푛푣(푇 − 푇0) Equation (24) 4 4 푞푟푎푑 = 휀휎[(푇 − 푇푎푏푠) −(푇0 − 푇푎푏푠) ] Equation (25) where T is the current temperature, T0 is the ambient temperature, Tabs is the absolute zero temperature, ε is the emissivity or the ability to emit thermal radiation and σ is the Stefan- Boltzmann constant (5.68 x 10 -8 J/K4m2s). The effect of radiation is greater at higher temperatures (close to the weld) and insignificant at low temperatures (far from the weld). According to Yang et al. [404] the emissivity of Ti-6Al-4V is almost independent of temperature at low temperatures up to 760°C with a value below 0.3, whereas, above 760°C it starts to oxidise and its emissivity increases with temperature to almost 1.0 above 1000°C. In contrast, the effect of convection is more significant at lower temperatures and less at higher temperatures. The calibrated values used for the heat transfer coefficients and radiation constants are listed in Table 15

Table 15 Calibrated values for heat transfer coefficients and radiation constants

Sample h (W/Km) ε AA 2024-T3 Butt joint 5 0.05 AA 2024-T3 T-joint 10 0.4 Ti-6Al-4V Butt joint 10 0.4 6.5 Mechanical Analysis A mechanical analysis was performed to investigate the effect of thermal expansion and contraction due to welding on the macroscopic residual stresses and distortions using the nodal temperature data imported from the thermal analysis as predefined field. The mechanical response of the material was calculated using infinitesimal strain theory as shown in Equation 26 assuming elastic-plastic behaviour with isotropic hardening law (von Mises rate independent deviatoric plasticity model).

푇표푡푎푙 푒 푝 푡ℎ 푐 푣푝 푡푝 휀푖푗 = 휀푖푗 + 휀푖푗 + 휀푖푗 + 휀푖푗 + 휀푖푗 + 휀푖푗 Equation (26)

푇표푡푎푙 푒 푝 푡ℎ where 휀푖푗 is the total strain, 휀푖푗 is the elastic strain, 휀푖푗 is the plastic strain, 휀푖푗 is the thermal 푐 푣푝 푡푝 strain, 휀푖푗 is the creep strain, 휀푖푗 is the viscoplastic strain and 휀푖푗 is the transformation plasticity strain. The total strain was composed of the above strain components. It was

203 assumed that the strain caused by creep, viscoplasticity and trip components are negligible and therefore, not included in the calculations. Since the viscoplastic effects were ignored, the yield stress was assumed to be independent of the strain rate and dependent upon plastic strain and temperature. According to Lindgren [371], welding simulations must account for elastic, plastic and thermal strains to calculate residual stresses. The thermal strains have the greatest influence on welding induced residual stresses and therefore, one of the most important material parameters for stress calculations is the coefficient of thermal expansion [366]. Lindgren [372] further suggested that that the final weld cooling phases are the important stages leading to residual stresses and distortion. In the case where a material undergoes phase transformation, the resulting volume changes can be included in the thermal strain component. However, the equivalent von Mises stress is not affected by the phase transformations [405]. Thermo-mechanical material properties used for welding simulations include elastic modulus, Poisson’s ratio, yield stress, density and thermal expansion coefficient. The elastic constants including the elastic modulus, yield limit and strain hardening were determined in isothermal uniaxial static tensile tests at different temperatures as shown in Figure 104 for AA 2024-T3 and Figure 106 for Ti-6Al-4V.

ABAQUS provides various metal plasticity model to define the true yield stress of the material as a function of true plastic strain and assumes no further work hardening beyond the last input data. In an isotropic hardening model, plastic flow causes uniform changes in the yield surface in all directions so the initial shape and orientation is maintained. In a linear kinematic hardening model, the yield surface is translated instead. The kinematic hardening model is usually used in situations where the material experiences cyclic behaviour such as low cycle fatigue which involves plastic flow and stress reversals, and the Bauschinger effect becomes relevant. Such effect causes the compressive yield strength to decrease by the same amount as the tensile yield strength increases after an initial tensile loading. Both isotropic and kinematic hardening models can be combined to give better predictions in cases involving cyclic loading but requires more detailed calibration. Mullins and Gunnars [406] conducted welding simulations using isotropic, kinematic and mixed hardening models and concluded that the isotropic hardening model produced simulation results which match experimental measurements the best and therefore, recommended for use in welding simulations. The mixed and kinematic hardening models underestimated the magnitude of residual stresses and were not recommended. According to Ogawa et al. [407], their results indicated that the choice of hardening model did not have a significant influence on welding residual stresses in 318 stainless steel girth welded pipes. Dong [408] simulated multi pass girth welding and determined that the kinematic hardening model is suitable for simulating multi pass welds due to the reversal of plasticity. When defining plasticity data in ABAQUS, in order to distinguish

204 differences in the plastic behaviour caused by the specimen geometry and the nature of the applied loads such compression and tension, true stress and strain were used which account for change in area during deformations, by converting nominal plastic material data from engineering to true values.

Welding thermal cycle is non-isothermal meaning that the use of isothermal elastoplastic data to define strain hardening during cooling at lower temperatures can be problematic because very low yield strength at high temperatures causes large plastic strains to develop which leads to artificial hardening and higher than expected residual stress values. To solve this problem, an annealing step was modelled by setting the equivalent plastic strain to zero above the specified annealing temperature so that the effect of prior hardening is lost. If the temperature at a material point falls below the annealing temperature, then it can work hardening again. By including the annealing effect, the residual stresses in the weld region can be reduced significantly.

According to Dye et al. [409] the lack of appropriate temperature dependent weld material properties does not substantially influence residual stresses and distortions since they only become significant when the material regains most of its strength after cooling. However, in the case of AA 2024-T3, the residual stresses obtained in the weld region can be much higher than the material’s yield strength. The welding heat causes dissolution or overaging of the strengthening precipitates in the FZ or the HAZ from the initial T3 temper. As a result, the yield stress of the weld metal is much less than the base metal. Owen et al. [410] simulated autogenous tungsten inert gas (TIG) welding of AA 2024-T3 and compared with experimental results from neutron and synchrotron measurements. They concluded that softening of the weld must be included to simulate resulting welding residual stress field. In order to account for the weld softening, the weld mechanical properties were used as field variables so that they can replace the base metal mechanical properties during cooling. Transition zones of intermediate mechanical properties between those of the base metal and the weld were also introduced to avoid non-gradual variations of mechanical properties in the weld region.

The addition of filler metal during welding was simulated using the user subroutine USDFLD to define solution dependent material properties as field variables. Two different techniques were used: quiet element, and element rebirth technique. In the quiet element (dummy) technique, the relevant elements were active all the time, whilst initially assigned dummy material properties and then reassigned the weld material properties as they start to cool so that their influence on the rest of the model can be minimised [382]. In the element rebirth technique, the relevant elements were assigned the same weld material properties and initially inactivated but they became activated as soon as the corresponding part of the weld was simulated. Lindgren et al. [411] simulated multi pass welding of a thick plate using both

205 methods and found that both methods produced very similar results but the element rebirth technique was more effective in terms of computational performance. A problem was found where decreasing the stiffness of the quiet elements excessively caused numerical problems due to ill conditioned stiffness matrices.

Mechanical boundary conditions were prescribed to fix the workpiece while it is being welded. Three different clamping conditions were considered as shown in Figure 111 to investigate the effects of different clamping conditions on welding residual stresses and distortions with the aim to control and minimise both of them to produce high strength, low stresses and dimensionally stable welded structures [412]. In the first case, the workpiece was welded without any restraints (free) except those to prevent rigid body motion. In the second case, the workpiece was completely fixed around all edges. In the last case, the boundary conditions were set to simulate the actual welding fixtures (clamp) with closer clamping to and larger restraints at the weld seam, used during welding experiments. It is a common practice to use fixtures to reduce welding distortions. However, distortion control is a difficult task because the welding fixtures strongly affect the residual stresses and distortions induced during welding due to the complexity of the welding process so they should be designed carefully.

Figure 111 Mechanical boundary conditions applied during welding and cooling for differernt simulations

The amount of restraint determines the magnitude of residual stress fields and distortions. Removal of the welding fixtures after cooling to room temperature partially releases the locked- in stresses or elastic strains by causing the workpiece to deform. In general, a low restraint causes large distortions but low residual stress fields. On the other hand, a high restraint

206 results in high residual stress fields but low distortions because it increases plastic deformation and reduces the remaining elastic strains which cause distortion after unclamping. This means that when over clamped, residual stresses build up and lead to weld cracking problems, whereas, when under clamped then there can be misalignment issues near the weld region which reduce dimensional stability of the welded part [413]. A successful residual stress and distortion control can lead to enhanced in-service structural performance of the welded structures such as fatigue and damage tolerance and also eliminate the need for time consuming and expensive post weld distortion corrections [414].

6.6 Results and Discussion on AA 2024-T3

6.6.1 Simulated Thermal Histories and Weld Seam Geometry Temperature distributions half way through welding AA 2024-T3 are shown in Figure 112 for thebutt welded specimens, Figure 113 for the single T-joint fillet welded specimen and Figure 114 for the triple T-joint fillet welded specimen. The figures indicate that the FE models have a very fine mesh density in and around the weld, where the temperature gradient is very steep and confined to a narrow region around the heat source at the location of keyhole formation reaching temperatures close to the boiling point of about 2500°C, and also elongated in the welding direction due to the high power density and low heat input characteristics of the moving fibre laser heat source. The welding heat quickly dissipates behind the heat source towards the lower temperature outer edges of the specimens due to very fast welding speed and elevated thermal conductivity values of the material at higher temperatures. It can be seen that the regions in white which are above the melting point of 638°C completely penetrate through the thickness for butt welded sheets, indicating a full penetration mode welding, whereas, only partially penetrate for fillet welded T-joint sheets.

Calibration of the welding temperature fields firstly involved matching the simulated weld pool geometry to the experimental macrograph of polished and chemically etched weld transverse cross-section. In the FE model, the molten zone or the FZ width was determined by the temperature contour above the liquidus temperature of 638°C and the HAZ width was determined by the temperature contour between the liquidus and the solidus temperature of 502°C. Figure 112, Figure 113 and Figure 114 show the simulated and experimental FZ side by side and it was found that the experimental fusion boundary and penetration depth are in good accordance with the simulated fusion boundary isotherms.

207 Temp ( C) 638 502 W47 P=4.9 kW, V=3.0 m/min 400 300 200 100 50 20

Temp ( C) 638 W99 P=2.9 kW, V=1.5 m/min 502 400 300 200 100 50 20

W102 P=4.9 kW, V=3.0 m/min, Vf=5.2 m/min

Temp ( C) W47 P=4.9 kW, V=3.0 m/min 2.1 mm 638 502 400 300 200 2.1 mm 100 50 W99 P=2.9 kW, V=1.5 m/min 20 3.5 mm

2.5 mm

W102 P=4.9 kW, V=3.0 m/min 3.2 mm

Vf=5.2 m/min

2.5 mm Figure 112 Temperature contours obtained from thermal analysis when half way through butt welding AA 2024-T3 and comparison of butt weld transverse cross-section geometry through thickness between FE model and experimental macrograph juxtaposed under three different welding conditions

The experimentally measured weld seams had the top and bottom widths as shown in Table 16. Small reinforcement and excessive penetration as observed in the actual macrograph were not modelled as they were assumed to have negligible influence on residual stresses and distortions. Table 16 also shows the calibrated heat source parameters re and ri as illustrated in Figure 109 used in the DFLUX subroutine for different welding conditions and

208 weld geometries, and the heat source efficiencies were determined to be in the range between 70 and 80%.

Table 16 Calibrated heat source parameters for welding AA 2024-T3 under various welding conditions

Top weld width Bottom weld width r r Sample name e i (mm) (mm) (mm) (mm) W47 2.1 2.1 0.1495 0.1485 W99 3.5 2.5 0.1905 0.1270 W102 3.2 2.5 0.1750 0.1395 Fillet 2.7 1.4 0.1750 0.1395

Dual beam Single beam Temp ( C) 638 502 400 300 200 100 50 20

Temp ( C) 638 502 400 300 200 100 50 20

Figure 113 Temperature contours obtained from thermal analysis half way through single T-joint fillet welding AA 2024-T3 and comparison of simulated and experimental weld transverse cross-section

Figure 113 shows two different welding modes for the single T-joint specimen, using either a single or dual laser beams. The actual specimen studied was only welded using a single laser beam due to problems associated with excessive heat input when welding using dual laser beam which resulted in poor weld quality and formation of welding defects. Thermocouple measurements were only taken for the single heat source case so the heat source parameters for the dual heat source FE model was based on the calibrated heat source for the single heat

209 source FE model, introduced to the opposite side of the stiffener as well to simulate simultaneous welding of both sides of the stiffener.

Temp ( C) 1st pass 638 502 400 300 200 100 50 20

2nd pass 3rd pass

Figure 114 Temperature contours obtained from thermal analysis half way through triple T-joint fillet welding AA 2024-T3 and comparison of simulated and experimental weld transverse cross-section

An additional calibration step involved matching the simulated curves to the temperature history recorded at various thermocouple positions by changing the thermal boundary conditions such as thermal convection and radiation to the surroundings and conduction to the worktop. Figure 115 shows that the experimental and simulated time-temperature curves at several locations match well in terms of heating rate, peak temperature and cooling rate. As the temperature reached significantly higher than the maximum operating temperature of the thermocouple in the FZ, the closest measurements were taken at some distance away from the weld centre line and the rest further away. The figure suggests that the heating and cooling speed are very fast during the welding process for all specimens, which was found to take less than a time period of 5 seconds only from heating to the peak temperature to cooling to less than 50°C. Since the specimens were all very thin, it was assumed that the temperature distribution through thickness is almost uniform regardless of the distance from the welding centreline. It was then decided that the thermal FE model experimentally validated through transient temperature and weld pool measurements became accurate enough to transfer the nodal temperature histories for stress analysis.

210 a) b) 700 500 W47 P=4.9 kW, V=3.0 m/min, f=+4 mm, W99 P=2.9 kW, V=1.5 m/min, f=+4 mm, 1.0 mm 3.0 mm 450 600 2.5 mm 4.0 mm 3.5 mm 400 7.5 mm 11.0 mm 500 5.0 mm 350

C) 6.0 mm TC1

° C) ° 300 400 TC1 TC2 TC2 250 TC3 300 TC3 TC4

200 Temperature Temperature ( Temperature( TC4 200 150 TC5 100 100 50 0 0 0 10 20 30 40 50 0 10 20 30 40 50 Time (s) Time (s)

c) d) 500 800 W102 P=4.9 kW, V=3.0 m/min, f=+4 mm, 2.5 mm 1.0 mm 450 Vf=5.2 m/min 3.5 mm 700 2.0 mm 400 6.5 mm 3.0 mm 600 9.5 mm 350 4.0 mm

22.0 mm C)

C) TC1

500 300 TC1 TC2 250 TC2 400 TC3 TC3 200 TC4

300 Temperature( TC4 Temperature( 150 TC5 Fillet P=3.5 kW, V=1.8 m/min, f=+0 mm, 200 100 Vf=5.0 m/min 100 50

0 0 0 10 20 30 40 50 0 10 20 30 40 50 Time (s) Time (s)

Post weld heat treatment for thermal stress relieving is difficult and impractical for heat treatable aluminium alloys. Typical artificial ageing temperature are not sufficiently high to effectively achieve significant reduction of the yield strength to obtain the desired stress relief. On the other hand, temperatures greater than the aging temperature and high enough to relieve stresses of around 232°C [415] cause further precipitation and overaging to occur and therefore, seriously reduce the strength level and further increase welding distortions. This means that welded structures which have been heat treated and aged to T3 temper prior to welding cannot be subsequently heated to the stress relieving temperatures.

6.6.2 Simulated Residual Stresses and Out of Plane Displacements Welding induces highly inhomogeneous stresses which may be as high as the yield strength of the material in the weld as well as considerable distortions. Stresses transverse to the welding direction are the x stresses, those parallel to the welding direction are the y stresses and those normal to the welding direction are the z stresses. Non-uniform thermal expansion and contraction due to different temperature gradients during the heating and cooling

211 sequence of welding process results in tensile and compressive residual stress fields in and near the weld region. Variations in the magnitude and distribution of residual stress fields due to welding were examined in this investigation, taking into account the different material properties including those of the base metal, weld metal and filler metal, along with other parameters such as geometrical and welding process parameters. In general, it was determined that fibre laser welding AA 2024-T3 components led to a highly localised narrow region of tensile y residual stresses around the weld with lower tensile stress magnitude softened weld metal and the remaining regions with almost no residual stresses or negligible compressive stresses. The contributions from x and z were relatively small and therefore, may be disregarded both for butt welded and fillet welded joints. Moreover, fibre laser welding the AA 2024-T3 welded structures also led to minimum welding distortions, typically less than a few millimetres, unlike significant welding distortions often observed with conventional welding processes.

In the case of butt welded sheets, the predicted residual stresses were symmetric across the weld centre line in and around the weld. Figure 116 and Figure 109 show the influence of mechanical boundary conditions on the welding induced residual stress distributions. Both the x and z stresses are very small compared to the y stresses across the width of the welded sheets, less than around a peak value of ±30 MPa. Low magnitude tensile stresses were observed in the weld and compressive stresses adjacent to the weld. These low magnitude stresses approached a zero value after a short distance away from the weld centre line. It appeared that there was some variation in x stresses through the thickness where compressive stresses were observed on the top and bottom surfaces and tensile stresses were observed at the interior. Similarly, tensile residual stresses were observed again on the top and bottom surfaces and compressive stresses at the interior adjacent to the weld. On the other hand, negligible variation through thickness were observed for y and z stresses.

Predominantly high magnitude tensile residual stress fields were observed in the y direction around the weld with the peak value of tensile residual stress in the FZ/HAZ boundary slightly lower than the yield strength of base metal at room temperature. Note that the maximum y residual stress was not located at the weld centre line due to softening as discussed previously. The highly tensile residual stresses around the weld quickly dropped again and approached a zero value away from the weld which tended to be weakly compressive. As it can be seen from Figure 109, only very narrow region within approximately ±5 mm around the weld were affected by the localised heating during the welding process due to low heat input and high power density of the fibre laser heat source, and therefore, largely tensile residual stresses were observed very close to the weld and the stress level in the remaining regions were close to zero. Although it was initially expected that weld joints with higher degree of restraint result

212 in higher residual stresses, the results from Figure 109 indicate that it was not necessarily true in this case. No significant variation in residual stresses was observed for different constraint sets examined after the removal of the restraints for all conditions. Instead, almost the same distribution and magnitude was observed. Therefore, it can be concluded that residual stresses are weakly sensitive to the mechanical boundary conditions and the extent of variation shows no significance of different constraints in reduction of residual stress fields for the components studied in this investigation.

Clamp Fix Free S11 (MPa) Clamp 25 20 15 10 Free 5 0 -5 -10 Fix -15 -20

S22 (MPa) Clamp 240 210 180 150 Free 120 90 60 30 Fix 0 -30

S33 (MPa) Clamp 25 20 15 10 Free 5 0 -5 -10 Fix -15 -20 Figure 116 Contours of residual stress distribution on the weld transverse cross-section and top surface of butt welded joints in x, y and z directions through thickness under three different mechaincal boundary conditions (AA 2024-T3, butt joint)

213 a) 30

20 σx

-10 Transverse(MPa) stress Clamp

Stress Stress (MPa) -20 Fix Free -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) b) 250

σy 200

150

Clamp 100 Fix Free

Stress Stress (MPa) Longitudinal Longitudinal stress (MPa)

-50 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) c) 30

σz 20

-10

Normalstress (MPa) Stress Stress (MPa) Clamp -20 Fix Free -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

Figure 117 Residual stress distributions in a) x, b) y, and c) z directions under three different mechanical boundary conditions representing clamping fixtures, perfect constraints around the workpiece edges and welding without any fixtures (AA 2024-T3, butt joint)

214 Clamp Fix Free U3 (mm) 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00

Figure 118 Contours of out of plane displacements on the top surface of butt welded joints under three different mechaincal boundary conditions (AA 2024-T3, butt joint)

It is difficult to differentiate from the stress contours in Figure 116, the minimal influence of mechanical boundary conditions on the residual stress distributions but there was clearly a major influence on the out of plane displacements as observed from the top surface of the specimens. Figure 118 shows that clamping close to the weld region as in the actual experimental clamping fixtures led to a slight reduction in compressive x and z residual stresses in and around the weld but the effect was negligible. In addition, almost no difference can be found in y stresses with respect to mechanical boundary conditions. On the other hand, displacement contours in Figure 118 and displacement plots in Figure 119 show that welding distortion was closely related to the magnitude of residual stresses as well as the degree of joint restraint during the welding process. The correlation between distortion and mechanical boundary condition was such that higher restraint led to lower distortion and vice versa. The least amount of out of plane displacement was achieved by using clamping fixtures located close to the weld region, followed by perfect restraints around outer edges and finally without any restraints.

Comparison of cambering and angular distortions from experiments with predicted data is shown in Figure 111. For all clamping conditions, the out of plane displacement was symmetrical about the weld centre line and close to zero at the weld start and end positions, and the greatest on the outer edges at the mid-length of the specimen. The predicted maximum cambering distortion of more than 2.0 mm was observed without restraints, 1.4 mm when perfectly fixed around the outer edges and 1.3 mm when clamped close to the weld centre line compared to the experimentally measured maximum cambering distortion of 1.8 mm for the actual clamping fixtures. The FE results without fixtures were over predicted compared to the experimental data but under predicted with fixtures. A small difference was identified between the experimental and predicted values of around 0.4 mm or 22% difference for stronger restraints close to the weld region. Still, a similar trend in cambering distortion was

215 observed both with measured and simulated distortion. In the case of angular distortions, the same pattern as that of cambering distortion was observed, where the specimen welded free of restraints over predicted the maximum angular distortion of 2.1 mm compared to the experimentally measured value of 1.6 mm, whereas, the specimens welded with fixtures under predicted with a value of 1.5 mm for weaker restraints around the outer edges and 1.1 mm for the clamping fixtures. This meant that the minimum cambering and angular distortions were achieved with higher degree of restraints such as clamping close to the weld region. Although it was not possible to completely eliminate welding distortions, it could still be optimised by using the appropriate mechanical boundary conditions as shown above, and therefore, proved the significance of restraints to enhance the dimensional stability of welded structures. a) b) 2.5 2.5

CMM Clamp Fix Free 2.0 2

1.5 1.5

1.0 1

plane displacementplane (mm)

- Out Out 0.5 0.5

CMM Clamp Fix Free

0.0 0 -250 -200 -150 -100 -50 0 50 100 150 200 250 -200 -150 -100 -50 0 50 100 150 200

Longitudinal distance (mm) Distance from specimen centre (mm)

Out of plane displacement (mm) displacement of plane Out Out displacement of (mm) Out plane Y distance(mm) Distance from weld centre (mm)

The influence of welding process parameters such as welding speed, laser power, heat input and welding with or without filler metal was studied. Three different combinations of welding parameters were investigated. For the sample W47, a laser power of 4.5 kW and welding speed of 3.0 m/min were used without filler metal. For the sample W99, a laser power of 2.9 kW and welding speed of 1.5 m/min were used also without filler metal. Finally, for the sample W102, the same laser power and welding speed as those of W47 were used but with a 1.0 mm diameter AA 4043 filler metal at a feed rate of 5.2 m/min.

Figure 120 and Figure 121 show the residual stress distributions for different sets of welding parameters used in numerical simulations. It is evident that the magnitude and distribution of residual stresses were influenced by the welding parameters. For all residual stresses including x, y and z stresses, the same trends were obtained for different welding parameters. Again, both x and z stresses were very small even in the weld region of magnitude less than

216 ±30 MPa. Almost no difference was observed with the z stresses, whereas, the magnitude of compressive x stresses adjacent to the weld increased in the order of W47, W99 and W102. Specimen W99, welded using lower laser power and slower welding speed than W47, had a slightly higher total heat input per unit volume, which directly influenced the temperature distribution and consequently the residual stress distribution.

W47 W99 W102 W47 S11 (MPa) P=4.9 kW 25 V=3.0 m/min 20 15 W99 10 P=2.9 kW 5 V=1.5 m/min 0 -5 W102 -10 P=4.9 kW -15 V=3.0 m/min -20 Vf=5.2 m/min

S22 (MPa) W47 P=4.9 kW 240 V=3.0 m/min 210 180 150 W99 P=2.9 kW 120 V=1.5 m/min 90 60 W102 30 P=4.9 kW 0 V=3.0 m/min

-30 Vf=5.2 m/min

S33 (MPa) W47 P=4.9 kW 25 V=3.0 m/min 20 15 10 W99 P=2.9 kW 5 V=1.5 m/min 0 -5 -10 W102 P=4.9 kW -15 V=3.0 m/min -20 V =5.2 m/min f

Figure 120 Contours of residual stress distribution on the weld transverse cross-section and top surface of butt welded joints in x, y and z directions under three different sets of welding parameters (AA 2024-T3, butt joint) More obvious variations in the residual stress distributions were observed for the three welding process parameters in the y direction. The magnitude of tensile residual stress at the weld centre line was higher for W99 than W47 and W102 by around 50 MPa. Specimen W99, welded using a reduced laser power and welding speed or less heat input, led to a smaller amount of softening in the weld which in other words, the yield stress of the weld metal for W99 was larger than the other two cases. Therefore, the magnitude of the residual stresses was higher, equal to around 150 MPa compared to around 100 MPa for W47 and W102. Figure 121 shows that the tensile residual stresses in the weld increased with increasing distance from the weld centre line up to around 3 mm, at which the peak value close to 250 MPa, comparable to the room temperature yield strength of the base metal, was reached. The y residual stresses then quickly dropped to values close to zero within few millimetres and remained uniform across the entire width of the specimens. The width of the higher magnitude

217 stresses in and around the weld was the narrowest for W47 which also had the smallest weld width.

a) 30

σx 20

-10 Transverse stress(MPa) W47

Stress Stress (MPa) -20 W99 W102 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) b)

250 σy 200

150 W47 100 W99 W102

Stress Stress (MPa) Longitudinal Longitudinal stress (MPa)

-50 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) c) 30

σz 20

-10

Normalstress (MPa) Stress Stress (MPa)

W47 -20 W99 W102 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

Figure 121 Residual stress distributions in a) x, b) y, and c) z directions under three different sets of welding parameters (AA 2024-T3, butt joint)

218 In addition, the maximum y tensile residual stress magnitude of around 240 MPa located at the FZ/HAZ boundary was roughly 10 MPa lower than W99 and W102. The extent of the softening region was similar for all cases but a considerable difference was observed in the peak tensile regions adjacent to the weld. It can be seen that this region was the smallest for W47, wider for W102 and the widest for W99. This could be attributed to increased heat conduction and wider temperature distribution of W99 due to its slower welding speed. Meanwhile, specimens W99 and W102 had almost the same peak y tensile residual stress magnitude as well as the distribution. This showed that welding with filler metal also acted to increase the amount of heat absorbed into the workpiece during welding and effectively increased the overall energy input even though the same welding parameters as W47 were used, except the addition or absence of filler metal.

W47 W99 W102 U3 (mm) 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00

Figure 122 Contours of out of plane displacements on the top surface of butt welded sheets under three different sets of welding parameters (AA 2024-T3, butt joint)

Figure 122 shows the influence of welding parameters on the out of plane displacements. When comparing W47 and W99, it can be seen that welding distortion could be minimised by using less heat input or higher welding speed. However, since the difference in heat input between W47 and W99 was very small, both the cambering and angular distortions were less than 1.0 mm at most and had fairly similar distributions. The simulated cambering distortions were over predicted compared to the measured values for W47 and W99 but the simulated angular distortions were in good agreement with the experimental results. On the other hand, the simulated values of both angular and cambering distortions for W102 were under predicted compared to the experimental measurements. The maximum cambering and angular distortions of more than 1.5 mm for W102 were more than two times greater than those of W47 and W99. Therefore, the results presented in Figure 115 showed that welding with filler metal can increase welding distortions since it is correlated to the energy input and weld microstructure so using less filler material is expected to lower welding distortions. At the same time, it is important to weld with filler metal to reduce problems associated with welding defects

219 so a balance is required to control both the weld quality and distortions. In conclusion, welding parameters have an influence on the residual stress magnitude and distribution to a certain extent but the variation was relatively small and the results obtained using different welding parameters were in good agreement with each other. The main differences were the magnitude of tensile residual stresses in the softened weld region and welding distortions due to different heat inputs caused by changing welding parameters and welding with or without filler metal.

a) b) 3.0 3.0 W47 (CMM) W47 (CMM) W47 (FEA) W47 (FEA) W99 (CMM) W99 (FEA) W99 (CMM) W102 (CMM) W102 (FEA) 2.5 2.5 W99 (FEA) W102 (CMM) W102 (FEA) 2.0 2.0

1.5 1.5 plane displacementplane (mm)

plane displacementplane (mm) 1.0 1.0

- Out Out 0.5 0.5

0.0 0.0 -250 -200 -150 -100 -50 0 50 100 150 200 250 -200 -150 -100 -50 0 50 100 150 200

Distance along specimen length (mm) Distance from specimen centre (mm) Out of plane displacement displacement (mm) of plane Out Y distance(mm) displacement of Out plane (mm) Distance from weld centre (mm)

Figure 123 a) Cambering and b) angular out of plane displacements under three different sets of welding parameters (AA 2024-T3, butt joint)

Welding residual stresses and distortions in T-joint fillet welded sheets were simulated as shown in Figure 126, Figure 119 and Figure 129. The influence of using two different techniques to simulate the addition of filler material, inactive elements (rebirth) or dummy material properties, on residual stresses and distortions was investigated. Figure 124 shows that careful implementation of both methods resulted in almost the same results in residual stresses and distortions. The dummy method predicted only around 20 MPa higher peak y stress magnitude close to 300 MPa than the inactive elements method but almost no difference in x and z stresses. When compared to the peak y stresses in butt welded sheets, they were around 50 MPa higher in fillet welded sheets and closer to the yield strength of the material at room temperature. They were likely to be due to stress triaxiality where the specimen thickness was greater around the stiffened region. The minimum y stresses are at the specimen centre (0 mm). The y stress increased rapidly with increasing distance from the specimen centre and reached the maximum at around 3 mm position where the fillet welds were located. It then drops to below 0 MPa and becomes compressive in nature and gradually tended to zero near the end of the specimen. Both the x and z stresses were very small compared to the y stresses but still greater than in butt welded specimens.

220 A trend that is similar to that of the y stresses was observed with the x stresses, whereas, the opposite trend was observed for the z stresses, where the peak z stress was found at the centre and the minimum at the fillet weld positions. In the case of distortions, welding with stiffeners reduced the cambering distortions significantly compared to butt welded sheets by almost ¾ but provided no clear advantage in terms of angular distortions. The inactive elements method predicted slightly greater cambering distortions but smaller angular distortions than the dummy method as shown in Figure 124 and Figure 126. However, the difference was very small. While the figure also shows that both methods under predicted the experimentally measured distortions, they were only different by around 0.3-0.4 mm for cambering and 0.7-0.8 mm for angular distortions. Therefore, the predicted results were reasonable and the difference in stress and distortion levels between the two methods became insignificantly small. a) b) 300 300 S11 (rebirth) S11 (rebirth) S11 (dummy) S11 (dummy) 250 S22 (rebirth) 250 S22 (rebirth) S22 (dummy) S22 (dummy) 200 S33 (rebirth) 200 S33 (rebirth) S33 (dummy) S33 (dummy) 150 150

100 100 Stress Stress (MPa) 50 Stress (MPa) 50

0 0

-50 -50

-100 -100 -120-100 -80 -60 -40 -20 0 20 40 60 80 100 120 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from specimen centre (mm) Distance from specimen centre (mm) c) d) 0.6 3.0 CMM Rebirth Dummy CMM Rebirth 0.5 2.5 Dummy

0.4 2.0

0.3 1.5 plane displacementplane (mm)

0.2 displacementplane (mm) 1.0

- Out 0.1 Out 0.5

0.0 0.0 -150 -100 -50 0 50 100 150 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 Distance along specimen length (mm) Distance from specimen centre (mm)

221 Residual stresses and distortions under three different mechanical boundary conditions were also investigated for the single T-joint fillet welded sheet and are presented in Figure 125 and Figure 126. It was found that unlike butt welded sheets, changing the clamping condition had some noticeable effect on the residual stress distributions as well as distortions. Figure 125 shows that increasing the restraints led to higher magnitude of residual stresses but lower distortions. The peak stresses in all three directions were obtained where the mechanical boundary conditions were set to resemble the actual clamping fixtures, followed by clamping around the outer edges and finally without any restraints, with the maximum y stress of around 280, 270 and 260 MPa, respectively. However, the lower magnitude y stresses at the weld centre line remained relatively constant regardless of the set boundary conditions. On the other hand, increasing the restraints caused more compressive x stresses and higher tensile z stresses at the weld centre line, and the opposite at the fusion boundary. The stress levels far from the weld zone remained not only constant but also uniform over the entire length regardless of the boundary condition. a) b) 300 300 S11 (fix) S11 (fix) S11 (free) S11 (free) 250 S11 (clamp) 250 S11 (clamp) S22 (fix) S22 (fix) 200 S22 (free) 200 S22 (free) S22 (clamp) S22 (clamp) S33 (fix) S33 (fix) 150 S33 (free) 150 S33 (free) S33 (clamp) S33 (clamp)

100 100 Stress Stress (MPa) 50 Stress (MPa) 50

0 0

-50 -50

-100 -100 -120-100 -80 -60 -40 -20 0 20 40 60 80 100 120 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from specimen centre (mm) Distance from specimen centre (mm) c) d) 0.6 6 CMM CMM Clamp Fix Free Clamp 0.5 5 Fix Free

0.4 4

0.3 3 plane displacement (mm)plane 0.2 displacementplane (mm)

- 2

- Out 0.1 Out 1

0.0 0 -150 -100 -50 0 50 100 150 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 Distance along specimen length (mm) Distance from specimen centre (mm) Figure 125 Residual stress distributions a) across the entire width and b) in the weld region in all three principal directions, and g) cambering and h) angular distortions under three differernt mechanical boundary conditions for a single T-joint fillet welded specimens (AA 2024-T3, T-joint)

222 The y residual stresses significantly affected the distortion pattern of the T-joint. Figure 126 illustrates the variations in the out of plane displacements on the top surface of the single T- joint fillet welded specimen under three different mechanical boundary conditions. It was clear that increasing the restraints resulted in lower out of plane distortions. Figure 125 shows that angular distortions were the lowest and in closest agreement with experimental values when using the clamping fixtures close to the weld region with only around 0.6 mm difference at most, followed by perfect restraints on the outer edges of around 1.6 mm difference, and the greatest without fixtures of around 3.0 mm difference from the maximum CMM value. On the other hand, relatively small difference in cambering distortions were found among difference clamping conditions, especially between the actual clamping fixtures and perfect restrains on the outer edges of only 0.05 mm difference in the peak cambering distortion values. The two clamping conditions under predicted the experimentally measured cambering distortions, while those without fixtures were closer to the real values. Still, the greatest difference between the measured and simulated cambering distortions was less than 0.3 mm so the numerical results were reasonable. In addition, the effect of using inactive elements or dummy material properties under these three mechanical boundary conditions was studied. Figure 126 shows that a small difference was found between the two methods in terms of the width of the distorted regions but the overall trend was very similar.

Clamp Fix Free

(mm)

Dummy Element rebirth Element

Figure 126 Contours of out of plane displacements on the top surface of single T-joint fillet welded specimens under three different mechanical boundary condition (AA 2024-T3, T-joint)

223 Figure 127 and Figure 120 illustrate the difference in stress and displacement between using a single laser beam to weld one side at a time and dual beams to weld both sides simultaneously. It can be seen that a relatively high levels of x, y and z stresses were observed with the dual beam specimen in the weld region and also wider stress distribution. The region where x and z stresses were weakly in compression for the single beam specimen was in tension for the dual beam specimen over a wider area. In the case of the single beam specimen, the peak y stresses were located at the root of fillet welds on both sides of around 280 MPa and much lower at the centre of the specimen of around 120 MPa. On the other hand, in the case of the dual beam specimen, the maximum y stress level attained was higher and wider than that of the single beam specimen of around 350 MPa and the central region was also largely tensile in nature. The results in this case showed almost no dip in the middle but instead stresses as high as 330 MPa. The same was true for x residual stresses where there was no large compressive dip of around -100 MPa at the centre for the dual beam specimen, as found in the single beam specimen, but instead reached around 30 MPa in the middle. Such behaviour was caused by the central region heating up to higher temperatures when using dual beams compared to a single beam. It therefore, led to a greater tensile stress states at the centre than when less affected by the welding heat.

The welding mode had a greater effect on the magnitude and distribution of z stresses. It was observed that the compressive fillet weld regions of the single beam specimen become tensile over a much wider area in the dual beam specimen and the peak tensile z stresses at the centre increased in magnitude from around 50 MPa to as high as 150 MPa. This meant that the z stress components were considered to have a stronger influence on the residual stresses in dual beam welding mode. Figure 127 and Figure 120 also show the out of plane displacements for the two welding modes. It can be seen that there was almost no difference in the cambering distortions but the angular distortions were greater in the single beam specimen than the dual beam specimen, and in closer agreement with the CMM values. Increased amount of heat input supplied in the case of dual beam mode led to wider residual stress distributions and reduced displacement magnitude by approximately 0.4 mm.

224 Single beam

S11 (MPa) S33 (MPa) 400 400 350 350 300 300 250 250 200 200 150 150 100 100 50 50 0 0 -50 -50 -100 -100

S22 (MPa) 400 350 300 250 200 150 100 50 0 -50 -100

S22 (MPa) U3 (mm) 400 0.20 350 0.00 300 -0.20 250 -0.40 200 -0.60 150 -0.80 100 -1.00 50 -1.20 0 -50 -100

Dual beam S11 (MPa) S33 (MPa) 400 400 350 350 300 300 250 250 200 200 150 150 100 100 50 50 0 0 -50 -50 -100 -100

S22 (MPa) 400 350 300 250 200 150 100 50 0 -50 -100

S22 (MPa) U3 (mm) 400 0.20 350 0.00 300 -0.20 250 -0.40 200 -0.60 150 -0.80 100 -1.00 50 -1.20 0 -50 -100

Figure 127 Contours of residual stress distribution and out of plane displacement on the weld transverse cross-section and top surface of a single T-joint fillet welded specimens in x, y and z directions welded using either single or dual laser beams (AA 2024-T3, T-joint)

225 a) b)

400 400 S11 (single) S11 (single) 350 S11 (double) 350 S11 (double) S22 (single) S22 (single) 300 S22 (double) 300 S22 (double) S33 (single) S33 (single) 250 250

200 200

150 150

100 100

Stress Stress (MPa) Stress Stress (MPa) 50 50

0 0

-50 -50

-100 -100 -120-100 -80 -60 -40 -20 0 20 40 60 80 100 120 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from specimen centre (mm) Distance from specimen centre (mm)

c) d)

0.6 3.0 CMM CMM Single Dual Single 0.5 2.5 Dual

0.4 2.0

0.3 1.5 plane displacementplane (mm)

0.2 displacementplane (mm) 1.0

- Out 0.1 Out 0.5

0.0 0.0 -150 -100 -50 0 50 100 150 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 Distance along specimen length (mm) Distance from specimen centre (mm)

The actual specimen was welded without restraints and so the simulated cambering distortions without fixtures were in good agreement with the CMM values, whereas, almost no cambering distortions were predicted when welded with fixtures. The predicted angular distortions when modelled without fixtures followed the same trend as those of the CMM distortions but under predicted by almost 3 cm at the edges. However, considering the size of the specimen, such difference was acceptable.

226 Constrained

S11 (MPa) S33 (MPa) 400 400 350 350 300 300 250 250 200 200 150 150 100 100 50 50 0 0 -50 -50 -100 -100

S22 (MPa) 400 350 300 250 200 150 100 50 0 -50 -100

S22 (MPa) U3 (mm) 400 5.0 350 4.5 300 4.0 250 3.5 200 3.0 150 2.5 100 2.0 50 1.5 0 1.0 -50 0.5 -100 0.0

Unconstrained

S11 (MPa) S33 (MPa) 400 400 350 350 300 300 250 250 200 200 150 150 100 100 50 50 0 0 -50 -50 -100 -100

S22 (MPa) 400 350 300 250 200 150 100 50 0 -50 -100

S22 (MPa) U3 (mm) 400 5.0 350 4.5 300 4.0 250 3.5 200 3.0 150 2.5 100 2.0 50 1.5 0 1.0 -50 0.5 -100 0.0

227 a) b)

350 S11 (free) 350 S11 (fix) S11 (free) 300 S22 (free) 300 S11 (fix) S22 (fix) S22 (free) 250 S33 (free) 250 S22 (fix) S33 (fix) S33 (free) S33 (fix) 200 200

150 150

100 100 Stress Stress (MPa) Stress Stress (MPa) 50 50

0 0

-50 -50

-100 -100 -200 -100 0 100 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from specimen centre (mm) Distance from specimen centre (mm) c) d) 0.6 7 CMM CMM Free Free 6 0.5 Fix Fix 5

0.4 4

0.3 3

2 plane displacementplane (mm)

plane displacementplane (mm) 0.2

of -

- 1

Out Out 0.1 0

0.0 -1 -200 -150 -100 -50 0 50 100 150 200 -200 -150 -100 -50 0 50 100 150 200 Longitudinal distance (mm) Transverse distance (mm)

6.7.1 Simulated Thermal Histories and Weld Seam Geometry The temperature distribution half way through welding the Ti-6Al-4V workpiece is shown in Figure 131. The FE model shows a fine mesh density in and around the weld, where the temperature gradient is very steep and confined to a narrow region around the heat source at the location of keyhole formation, reaching temperatures close to the boiling point and also elongated in the welding direction because of the high power density and low heat input characteristics of the moving fibre laser heat source. The welding heat quickly dissipates behind the heat source towards the lower temperature outer edges of the sheet due to very high welding speed and elevated thermal conductivity values of the material at higher temperatures. It can be seen that the regions in white which are above the liquidus

228 temperature of 1650°C completely penetrate through the thickness of the sheet, indicating a full penetration mode welding.

Figure 131 Temperature contours obtained from thermal analysis half way through welding Ti-6Al-4V

Calibration of the welding temperature fields involved matching the simulated weld pool geometry to the experimental macrograph of polished and chemically etched weld transverse cross-section. In the FE model, the FZ width was determined by the temperature contour above the liquidus temperature of 1650°C and the HAZ width was determined by the temperature contour between the liquidus and the solidus temperature of 1600°C. It was difficult to precisely match the HAZ width due to very narrow gap between the liquidus and the solidus but still a good agreement was made. Figure 132 shows the simulated and experimental FZ side by side and it shows that the experimental fusion boundary and penetration depth were in good accordance with the simulated fusion boundary isotherm. The experimentally measured weld seam had a top width of 2.8 mm and bottom width of 2.35 mm. Small reinforcement and excessive penetration as observed in the actual macrograph were not modelled as they were assumed to have negligible influence on residual stresses and distortions. The heat source after calibration had a laser power of 3 kW, heating efficiency of

80%, welding speed of 3 m/min, re equal to 0.4 mm and ri equal to 0.3 mm.

229 2.80 mm Temp (°C)

2.35 mm Figure 132 Comparison of weld transverse cross-section geometry through thickness between FE model and experimental macrograph juxtaposed (Ti-6Al-4V, butt joint)

Further calibration on the welding temperature fields involved matching the simulated curves to the temperature history recorded at various thermocouple positions by changing the thermal boundary conditions such as thermal convection and radiation to the surroundings and conduction to the worktop. Figure 133 shows that the experimental and simulated time- temperature curves matched well in terms of heating rate, peak temperature and cooling rate. As the temperature reached significantly higher than the maximum operating temperature of the thermocouple in the FZ, the closest measurement was taken at a distance of 3 mm away from the weld centre line and a few more further away. It can be seen that the simulated calibrated curves presented marginally faster cooling rate than the experimental curves but were in very good agreement overall after several iterations. The cooling rate at distances greater than 4 mm away from the weld centre line showed significantly reduced peak temperature while still reasonably maintaining similar heating and cooling rates compared to those closer to the weld. This meant that the area affected by the welding heat source was mostly within less than 5 mm. It was decided that the thermal FE model experimentally validated through transient temperature and weld pool measurements became accurate enough to transfer the nodal temperature histories for stress analysis.

700 3.0 mm

600 3.5 mm 3.8 mm 500 4.0 mm

5.0 mm

C) ° 400 TC1 TC2 300 TC3

Temperature( TC4 200 TC5

100

0 0 10 20 30 40 50 Time (s)

Figure 133 Numerically calculated thermal histories calibrated using thermocouple measurements at various distances from the weld centre line (Ti-6Al-4V, butt joint)

230 6.7.2 Simulated Residual Stresses and Out of Plane Displacements Welding induces highly inhomogeneous stresses which may be as high as the yield strength of the material in the weld as well as considerable distortions. Figure 134 shows the stress contour plots in x, y and z directions on the top surface and weld cross-section at the end of cooling and relaxation, where x stress perpendicular to the welding direction and y stress along the welding direction are the in plane stresses, and the z stress is the out of plane stress. The y stress was found to be the maximum in and near the weld region in tension with weakly compressive stress field in the rest of the workpiece far from the weld. The x and z stresses were the maximum and tensile in nature in the FZ, compressive near the weld and almost stress free further away from the weld. All three principal stresses were symmetrical due to symmetry across the weld centre line.

Figure 134 and Figure 135 also show the effect of mechanical boundary conditions on residual stresses. It had very small effect on residual stresses, where all three principal stresses were only weakly sensitive s. No restraint or weaker restraint around the edges exhibited slightly higher residual stresses whereas, comparatively stronger rigid restraint such as the clamping fixture used for real welding experiments produced slightly lower residual stresses. The maximum y stress reached under all conditions were approximately in the range between 750 and 780 MPa, comparable to the yield strength but still not as high. It was determined that the amount of variations in three residual stress fields observed for different clamping conditions simulated in this investigation indicated no significance of constraints in reduction of residual stress fields. Clamp Fix Free S11 (MPa) Clamp 80 60 40 20 Free 0 -20 -40 -60 -80 Fix -100

S22 (MPa) Clamp 800 700 600 500 400 Free 300 200 100 0 Fix -100

S33 (MPa) Clamp 40 30 20 10 Free 0 -10 -20 -30 -40 Fix -50

Figure 134 Residual stress distributions on the weld cross-section (20 mm wide) and top surface in prinicipal directions for three differernt mechanical boundary conditions (Ti-6Al-4V, butt joint) 231 a) b) 800 80 S11 (Clamp) 700 S22 (Clamp) S33 (Clamp) 60 S11 (Fix) 600 S22 (Fix) S33 (Fix) 40 500 S11 (Free) S22 (Free) S33 (Free) 400 20

300 0 Stress(MPa) 200

-20 Transversestress (MPa) 100 Clamp -40 Fix 0 Free -100 -60 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) c) d) 800 80

700 60 600 40 500

400 Clamp 20 Fix 300 Free 0 200

Normal stress(MPa) -20 Longitudinal Longitudinal stress (MPa) 100 Clamp -40 Fix 0 Free -100 -60 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Figure 135 Residual stress distributions a) across the entire width in all three principal directions, near the weld in b) x, c) y, and d) z directions (Ti-6Al-4V, butt joint)

Figure 136 and Figure 137 show the effect of mechanical boundary conditions on the magnitude of cambering and angular out of plane distortions. For all clamping conditions, the out of plane displacement was symmetrical about the weld centre line and close to zero at the weld start and end positions, and the greatest on the outer edges at the mid-length of the specimen. Simulated displacement results were all greater than the experimental results from CMM measurements especially for cambering distortions, whereas, angular distortions were predicted better. Highest values of out of plane deformations were observed when free of any restraint, moderate when perfectly fixed around the outer edges and the lowest when most rigidly restraint close to the weld region. As expected, the clamping model predicted the out of plane displacements closest to the CMM values. Although it was not possible to completely eliminate welding distortions, it could still be optimised by using the appropriate mechanical boundary conditions as shown above, and therefore, proved the significance of restraints to enhance the dimensional stability of welded structures.

232 U3 (mm) 9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Clamp Fix Free Figure 136 Out of plane displacements after cooling down to room temperature and removing any restraints for three different mechanical boundary conditions (Ti-6Al-4V, butt joint) a) b) 6 6 y=250 (CMM) y=250 (Fix) y=250 (Free) y=250 (Clamp) 5 5

4 4

3 3 plane displacementplane (mm)

plane displacementplane (mm) 2 2

Out Out 1 1 x=0 (CMM) x=0 (Fix) x=0 (Free) x=0 (Clamp) 0 0 -250 -200 -150 -100 -50 0 50 100 150 200 250 -200 -150 -100 -50 0 50 100 150 200

Distance along specimen length (mm) Distance from specimen centre (mm)

Out of plane displacement displacement (mm) of plane Out Out of plane displacement displacement (mm) of plane Out Y distance(mm) Distance from weld centre (mm)

Figure 137 a) Cambering and b) angular out of plane displacements for three different mechanical boundary conditions (Ti-6Al-4V, butt joint)

6.7.3 Modelling Solid State Phase Transformations in Ti-6Al-4V Welds consist of a complex heterogeneous microstructure within a small volume due to very steep temperature gradients during welding. Solid state phase transformation may occur during cooling and the associated volume change and phase specific material properties can have a significant influence on welding residual stresses and distortions. In many weld models, the influence of phase transformation is neglected. However, according to Ferro et al. [380] such effects are not negligible in the case of bead on plate laser welding. Leblond [416] stated that it is important to include the solid state phase transformations in the model in order to obtain accurate deformations but not for residual stresses.

Several researchers studied solid state phase transformation in steel due to austenitic and martensitic transformation, and transformation plasticity. Yaghi et al. [417–419] simulated tungsten inert gas and manual metal arc welding P91 steel pipe with solid state phase

233 transformation taking into account transformation plasticity, volume changes calculated by using only a single value for volumetric change strain as a function of phase fraction and associated changes in yield strength due to austenitic transformation from pearlite and ferrite on heating and martensitic transformation during cooling as a function of transformation start and end temperatures but not considered cooling rate. Similarly, Lee and Chang [420] studied solid state phase transformation in flux cored arc welded carbon steel, and Dean and Hidekazu [421] modelled multipass 9Cr-1Mo steel pipe gas tungsten arc welding (GTAW). It was determined that the volume change due to martensite transformation has a significant influence on welding residual stress. It not only changed the magnitude of the residual stress, but also inverted the sign of residual stresses in the weld zone. Their FE results when taking into account the volume change due to martensite transformation were in good agreement with the strain gauge experimental results. Taljat et al. [422] also discovered that austenite to martensite transformation relieved high tensile residual stresses in the weld in gas tungsten arc welded HY-100 steel and the FE results matched ND measurements.

Transformation induced volumetric strains in steels are caused by phase work hardening due to differences between the specific volumes and thermal expansion coefficients of phases, and promote reorientation of new phases [423]. On the other hand, this is not the case with pure titanium, because the specific volumes of α and β unit cells are similar. This means that is there is almost no change in crystallographic orientation during α to β allotropic transformation and therefore, the phase work hardening effect is not significant and only small internal stresses are produced. In the case of Ti-6Al-4V, the transformation occurs in some temperature range rather than in a particular point during and due to the presence of some β phase in the initial base metal microstructure, so the transformation proceeds from α + β to β via the increase of β phase volume by diffusional migration of the α/β interface [423]. Therefore, the volume effect of transformations in Ti-6Al-4V was modelled by taking into account changes of specific volumes of different phases determined from the lattice parameters of these phases. The linear coefficients of thermal expansion for α and β phases as a function of temperature were obtained from JMatPro [424] as shown in Figure 138 in order to account for internal stresses caused by the differences in the coefficients in calculations.

234 a) b)

30 37.0 0.00 L

α 25 36.5

-0.01 ) 20 3 36.0

-0.02

/K) 6 - 15 35.5 Beta

(10 Alpha -0.03 Alpha to Beta

10 35.0 Volumetric strain Unit cell volume (Å

-0.04

5 Beta 34.5 Linear Linear coefficient of thermal expansion, Alpha

0 34.0 -0.05 0 500 1000 1500 2000 0 200 400 600 800 1000 Temperature ( C) Temperature ( C)

Figure 138 a) Linear coefficient of thermal expansion and b) unit cell volume for each phase, and transformation induced volumetric strains as a function of temperature [425]

The unit cell volume for each phase was determined from the real time in-situ synchrotron X- ray diffraction measurements of the lattice parameters of α and β phases during heating of Ti- 6Al-4V by Elmer et al. [425]. The lattice parameters were measured from the bcc {110} reflection for β phase and hcp {101} reflection for α phase and the volume of α phase was

2 3 calculated using a0 csin60° assuming c/a = 1.5963 and a0 for β phase as shown in Figure 138. The volumetric strains during phase transformation due to the differences in unit cell volumes between these two phases was therefore, calculated by multiplying the volume fraction of each phase in the microstructure by its respective unit cell volume as a function of temperature during welding.

Few researchers studied the kinetics of phase transformations in Ti-6Al-4V but were mostly case specific because they are known to vary with the chemical composition, the temperature history and the initial phase morphology prior to transformations. Shah et al. [426,427] initially modelled analytically various phase transformations during welding thermal cycle using modified Rosenthal equations in the heat affected zone in terms of heat input and plate thickness. Malinov et al. [428,429] studied the phase transformation kinetics under isothermal condition using resistivity technique and continuous cooling conditions using differential scanning calorimetry at different cooling rates using the Johnson Mehl Avrami and Kolmogorov (JMAK) equation [430–432] and the additivity rule to determine the fraction of β transformed to α. Ahmed and Rack [282] studied the effects of cooling rate from above the beta transus on phase transformation during cooling and determined that martensitic formation forms at high cooling rates above 410 K/s, while Gil Mur et al. [331] studied the decomposition of martensite into α and β during tempering heat treatment and observed fully martensitic structure at much lower rates. Kelly and Kampe [433,434], Crespo et al. [435] and Murgau et al. [436] modelled microstructural evolution in Ti-6Al-4V during laser metal

235 deposition process and Fan et al. [329] numerically investigated the effect of phase transformations on laser forming of Ti–6Al–4V using JMA kinetic parameters for isothermal transformation from Malinov et al. [428] but without applying additivity principles to diffusion controlled transformation. Elmer et al. [281] measured experimentally using synchrotron X-ray diffraction technique the transformation kinetics during gas tungsten arc welding (GTAW) of Ti-6Al-4V. Longuet et al. [437] developed a general multiphase model for Ti-6Al-4V and applied to direct laser fabrication and laser welding processes.

In this investigation, FORTRAN user subroutines were developed based on time temperature transformation (TTT) diagrams to include the effect of solid state phase transformations in Ti- 6Al-4V during welding. The transformed volume fraction of different microstructural phases was calculated in the thermal analysis as internal state variables using USDFLD as a function of cooling rate and peak temperature, and the volumetric change due to temperature variations and also phase transformations were calculated in the mechanical analysis using UEXPAN by mimicking as the thermal strain due to thermal expansion as discussed in Section 6.5.

During the heating stage, Ti-6Al-4V undergoes a rapid allotropic diffusion controlled transformation from α to β phase which involve nucleation and growth of β and dissolution of α by movement of the α/β interface due to the transport of beta stabiliser across the interface [423]. The volume fraction of β reaches 100% above the β transus temperature and remains stable up to the melting temperature. Since is β an intragranular phase, it is difficult to evaluate its volume fraction accurately. As Equation 27 shows, it was assumed that β growth increases by the equivalent amount of α dissolution and that the transformed β reaches its equilibrium fraction instantaneously during heating and follow the beta equilibrium curve in Figure 139. A simplified single α phase was assumed to avoid differentiating variants of phase morphologies such as globular, Widmanstätten, basket weave and grain boundary α phases. As Equation 28 shows, a single α phase was calculated by subtracting the current β phase fraction from the equilibrium β phase fraction.

푒푞 푓훽(푡1, 푇1) = 푓훽 (푇1) Equation (27)

푒푞 ∆푓훽 = −∆푓훼 = 푓훽 (푇1) − 푓훽(푡0, 푇0) Equation (28)

Figure 139 shows the equilibrium fraction of α and β phases determined from various models in the literature. A similar trend can be found in all models except for the JMatPro model which predicts the volume fraction of α as 0% and 100% for β at low temperatures, whereas, other models predict around 10% retained β when cooled to room temperature. Since some β phase was present in the actual base metal microstructure, the Castro model was used in the numerical simulations to predict the equilibrium phase fractions.

236 1.0 1.0 β (JMATPRO)

β (Avrami) α 0.8 β 0.8 β (Castro) β (Charles)

0.6 0.6

0.4 0.4

α (JMATPRO)

α (Avrami) Phase Phase fraction beta of phase, Phase fraction ofalpha phase, 0.2 0.2 α (Castro) α (Charles) 0.0 0.0 0 500 1000 1500 0 500 1000 1500 Temperature ( C) Temperature ( C)

Figure 139 Modelled equilibrium phase fraction of α and β phases as a function of temperature [424,436,438,439]

During the cooling stage, the cooling rate determines the transformation kinetics and the resultant phases formed. For slow cooling rates less than around 20 K/s at temperatures below the beta transus, the β decomposes into α via a nucleation and diffusion controlled reaction [282,436,440]. The diffusive transformation of the β phase into the α phase which occurs at a constant temperature can described using the Johnson Mehl Avrami and Kolmogorov (JMAK) law [430–432] as shown in Equation 29.

푛훽→훼 푒푞 푓훼(푡, 푇) = [1 − exp (−푘훽→훼(푇)푡 ]푓훼 (푇) Equation (29)

푒푞 where is 푓훼 the fraction of the α phase at time 푡, 푓훼 is the equilibrium fraction of the α phase, and 푘훽→훼 and 푛훽→훼 are the temperature dependent and independent JMAK parameters that define the kinetics of the β to α transformation, determined from available time temperature transformation (TTT) start and end curves from the literature for this transformation [441] as shown in Figure 140.

As it can be seen from Figure 140, there are considerable variations in the TTT curves [429,433,436,442] so the derived JMAK parameters depend on the selected of TTT curves.

The JMAK parameters, 푘훽→훼and 푛훽→훼 = 2.5 used in this investigation were based on the TTT diagram for Ti-6Al-4V in Kelly [443] obtained by fitting the modelled JMatPro TTT curves as shown in Figure 140..

237 1000 Tβ 900

800

700

600

500 1% and 95% α-w [449] Temperature ( Temperature 1% and 95% α-gb [449] 1% and 50% α [434] 400 5% and 95% α [435] 50% α [434] 5% and 95% α [442] 300 Start and end α [442] Start α [442] Start and end α [442] 200 1% and 95% α combined [449] 0.1 1 10 100 1000 10000 1000001000000 Time (s)

Figure 140 Time temperature transformation (TTT) diagrams from the literature determined experimentally and calculated using JMatPro [428,429,436,443]

0.014 n = 2.5

0.012

0.010

0.008 k(T) 0.006

0.004

0.002

0.000 200 400 600 800 1000 1200 Temperature ( C)

Figure 141 Kinetic parameters for diffusion controlled transformations derived from JMatPro TTT diagrams using the interpolating function defined by Kelly for 1 and 95% of transformation [443]

The JMAK models for both dissolution, and nucleation and growth of α phase are based on kinetic parameters derived for isothermal transformation. In order to extend the application of the isothermal models to non-isothermal phase transformations during welding, Scheil’s additivity rule was used [444]. The additivity rule approximates an arbitrary continuous temperature variation, in this case the total time to reach a specific stage of transformation, as a sum of small incremental isothermal time steps connected by instantaneous temperature change [436,442,445]. The phase fractions at any time and temperature steps depend on the values from the previous steps and can be expressed as shown in Equation 30. 푡 푑푡 ∫ = 1 Equation (30) 0 푡푓(푇)

238 where 푡푓(푇) is the isothermal time to stage f and t is the time to f for non-isothermal reaction. The modified JMAK equation is shown in Equation 31.

푓 (푡 , 푇 ) = [1 − exp (−푘 (푇 )(푡∗ + 푡 − 푡 )푛훽→훼(푇1)](푓 (푡 , 푇 ) 훼 1 1 훽→훼 1 0 1 0 훽 0 0 Equation (31) ( ) 푒푞 ( ) + 푓훼푚 푡0, 푇0 )((푓훼 ( 푇1 )

∗ where 푡0 is a fictive time required to reach 푓훼(푡0, 푇0) during an isothermal transformation at 푒푞 temperature 푇1, 푓훼 is the equilibrium fraction of the α phase, which represents the fraction of the α phase. The expression for the fictive time is given in Equation 32. 1 푒푞 1 푓 (푇 ) 푛훽→훼(푇1) ∗ 훼 1 Equation (32) 푡0 = [ 푙푛 푒푞 ] 푘훽→훼(푇1) 푓훼 (푇1) − 푓훼(푡0, 푇0)

Very fast cooling from above the beta transus transforms β into another form of α phase thin needle like acicular martensite, α´ that is different from the equilibrium α, with high residual stresses due to relatively larger differences in the crystallographic orientation of adjacent lamella and interfaces with low coincidence [440]. The martensitic transformation is a diffusionless solid state transformation which causes very fast changes in the crystal lattice structure without rearranging the atoms. It can be defined by the Koistenen Marburger (KM) law as shown in Equation 33.

푓 ´(푡, 푇) = [1 − exp (−푘 ´(푀 − 푇)] 푓 (푡, 푇) 훼 훽→훼 푠 훽 Equation (33)

where 푓훼´ is the fraction of α´, 푓훽 is the fraction of 훽 and 푘훽→훼´ or γ is the material dependent

KM parameter and 푀푠 is the martensite start temperature. Figure 142 shows the equilibrium phase fraction of β and the rest consisting of α´ as a function of γ and 푀푠 determined using the

KM equation. The γ was chosen to be equal to 0.015 and the 푀푠 equal to 650°C. a) b)

1.0 1.0

TMs TMs Tβ TMs Tβ 0.8 0.8

α' (Ms= 575 C) α' (γ= 0.003) α' (Ms= 650 C) 0.6 0.6 α' (γ= 0.005) β (Ms= 575 C) α' (γ= 0.015) β (Ms= 650 C) β (γ= 0.003)

0.4 0.4 Phase Phase fraction Phase Phase fraction β (γ= 0.005) β (γ= 0.015)

0.2 0.2

0.0 0.0 0 200 400 600 800 1000 0 200 400 600 800 1000 Temperature ( C) Temperature ( C)

Figure 142 Phase fraction of β and α´calculated using the Koistenen Marburger model as a function of a)

푴풔 and b) γ

239 For fast cooling rates greater than 410°C/s, a fully martensitic microstructure is formed and diffusive transformation to α is suppressed according to Ahmed and Rack [282].

푓훼´(푡1, 푇1) = 푓훼´(푡0, 푇0) + ∆푓훼´ Equation (34) = [1 − exp (−푘훽→훼´(푀푠 − 푇)] (푓훽(푡0, 푇0) + 푓훼´(푡0, 푇0))

For moderate cooling rates between 20 and 410°C/s from above the beta transus leads to partial transformation of β to grain boundary massive 훼푚 and intragranular martensitic alpha

αʹ adjacent to the prior β grain boundary. It was assumed that 훼푚 includes both 훼푚and αʹ and also that both transformations are diffusional and martensitic.

( ) ( ) 푓훼푚 푡1, 푇1 = 푓훼푚 푡0, 푇0 + ∆푓훼푚 ( ) ( ) ( ) = [1 − exp (−푘훽→훼푚 푀푠 − 푇 ](푓훽 푡0, 푇0 + 푓훼푚 푡0, 푇0 Equation (35) 푒푞 − 푓훽 (푇1))

푒푞 where 푓훼푚 is the fraction of 훼푚 and 푓훽 is the equilibrium fraction of β. In this case, the current equilibrium fraction of β is subtracted from the β phase available for the transformation in order to prevent a full transformation to α´.

Post weld heat treatment (PWHT) is often used to reduce and redistribute the residual stresses that result from welding titanium and titanium alloys [446], without seriously affecting the mechanical properties such as strength or ductility. Stress relief enhances dimensional and structural stability as well as prevents problems such as stress corrosion cracking and loss of compressive yield strength. Unlike aluminium alloys, titanium alloys have high yield strength and elastic modulus at room temperature so post weld mechanical treatment such as stretching is less effective than PWHT due to significant springback. The PWHT process involves heating the workpiece to the heat treatment temperature, holding at that temperature for a specific time and then cooling down uniformly and slowly back to room temperature to avoid introducing additional thermal stresses due to temperature gradients. Stress relieving is a recovery/recrystallization process which occurs by increasing the temperature of the workpiece to provide additional energy that increase the rate of diffusion. The diffusion of atoms effectively removes or redistributes dislocations within the material and as a result, decreases the deformation resistance [447]. The extent of stress relaxation depends on the temperature and the soaking time. As can be seen from Figure 143, the yield strength of Ti- 6Al-4V decreases with increasing temperature due to dislocation activation. By uniformly heating the welded material, the yield limit is reduced and the stress in the material becomes greater than the yield strength at the treatment temperature. The material then plastically deforms by local creep and any residual stresses that exist in the material are reduced to the yield strength at the treatment temperature.

240 1200

1000

800

0.1/s 600 0.01/s 0.001/s

400 0.2% Proof 0.2% Proof stress(MPa)

200

0 0 200 400 600 800 1000 1200 1400 Temperature ( C)

Figure 143 Variation of 0.2% offset yield strength of Ti-6Al-4V with temperature at three different strain rates calculated from JMatPro

The welded Ti-6Al-4V sheet studied in this investigation was heat treated at 600°C for 6 hours.

In addition to stress relief, PWHT also causes partial recovery of massive αm and martensitic α ´ phases to α and β phases after heating and soaking at PWHT temperature for a long time by diffusional process and then decomposition of the previously transformed β into α via a nucleation and diffusion controlled reaction during cooling to room temperature until an equilibrium proportion of the β phase is reached. The final microstructure then mainly consists of αm and α, and a very small equilibrium fraction of β.

Figure 144 and Figure 145 show the results from numerical simulation on the variations of the volume phase fractions of α, β, and αm with welding thermal cycle and PWHT, at the weld centre and around the weld, respectively. The initial microstructure consisted of around 0.9 α and 1.0 β which are equivalent to the equilibrium volume fractions. As the temperature rapidly increases during heating, the α phase fraction quickly drops to 0 and β increases up to 1, which in reality should also become 0 above the melting temperature but for simplicity, it is maintained at 1 above the beta transus temperature. When the material starts to cool below the beta transus temperature, depending on the cooling rate, a different microstructure is produced either by diffusion or diffusionless transformation from the β phase. At the weld centre where the peak temperatures reached is very high and the cooling rate is fast, a diffusionless martensitic transformation occurs. It was found from the temperature history that the maximum cooling rate was around 400°C/s which is slightly less than the cooling rate of 410°C/s needed for a fully martensitic transformation. For that reason, the martensite phase fraction does not reach 1 but only up to around 0.9 and the remaining phase fraction consisted of massive alpha, αm and no β after cooling down to room temperature. Increasing the temperature again slowly to 600°C causes partial recovery of α and β by decomposition of martensite. The fraction of α increases to around 0.3, β to around 0.2 and martensite

241 decreases to around 0.5. Upon uniform cooling to room temperature, the fraction of α further increases by dissolution of β until equilibrium fraction of β is reached, while the fraction of martensite remains the same. The final microstructure is now composed of approximately 50% α and 50% martensite and negligible amount of β. a) b) 4000 700 1 1 0.9 3500 0.9 600 0.8 Temperature 3000 0.8 500 Beta 0.7

0.7

2500 Alpha 0.6 0.6 400 Temperature Martensite 2000 0.5 Beta 0.5 300

Alpha 0.4 Phase fractionPhase

1500 0.4 Phase fraction Temperature( Martensite Temperature( 0.3 200 0.3 1000 0.2 0.2 500 100 0.1 0.1

0 0 0 0 0 1 2 3 4 5 6 7 8 0 2 4 6 8 10 Time (s) Time (hrs)

β α αm 1.0 Start 0.9 0.8 Heating 0.7 0.6 Cooling 0.5 0.4 0.3 End 0.2 0.1 PWHT 0.0

Figure 145 Simulated weld transverse section (20 mm wide) showing the phase fractions of α, β and αm phases at different stages of welding and after PWHT at 600°C for 6 hours (Ti-6Al-4V, butt joint)

242 a) b) 800 80 S11 (PWHT) S11 700 S22 (PWHT) 60 S11 (εtrans) S33 (PWHT) S11 (PWHT) 600 S11 (εtrans, PWHT) S11 (εtrans, PWHT) S22 (εtrans, PWHT) 40 S33 (εtrans, PWHT) 500 S11 S22 20 400 S33 S11 (εtrans) 0 300 S22 (εtrans)

Stress Stress (MPa) -20

200

Stress Stress (MPa) Stress Stress (MPa)

Transverse (MPa) stress Transverse -40 100

0 -60 σx

-100 -80 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld centre (mm) c) d) 800 80 S22 S33 700 S22 (εtrans) 60 S33 (εtrans) σy S22 (PWHT) σz S33 (PWHT) 600 S22 (εtrans, PWHT) S33 (εtrans, PWHT) 40 500 20 400 0 300 -20

200 Normal Normal (MPa) stress

Longitudinal (MPa) stress Longitudinal -40 Stress Stress (MPa) 100 Stress (MPa)

0 -60

-100 -80 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld centre (mm) Distance from weld centre (mm) e) f) 8 4.0 CMM CMM Norm 7 3.5 Norm PWHT PWHT εtrans εtrans 6 εtrans+PWHT 3.0 εtrans+PWHT 5 2.5

4 2.0

3 1.5

plane plane displacement (mm)

of -

2 - 1.0

Out Out 1 0.5

0 0.0

-250 -200 -150 -100 -50 0 50 100 150 200 250 -200 -150 -100 -50 0 50 100 150 200 Out of plane displacement (mm) displacement of plane Out Distance along specimen length (mm) displacement (mm) of plane Out Distance from specimen centre (mm) Y distance(mm) Distance from weld centre (mm)

243 No Phase Phase Transformation Transformation S11 (MPa) 80 60 40 20 No Phase Transformation 0 -20 -40 -60 -80 -100 Phase Transformation

S22 (MPa) 800 700 600 500 No Phase Transformation 400 300 200 100 0 -100 Phase Transformation

S33 (MPa) 40 30 20 10 No Phase Transformation 0 -10 -20 -30 -40 -50 Phase Transformation Figure 147 Residual stress distribution in three principal stress directions after welding, on the weld transverse cross-section (20 mm wide) and top surface, with phase transformation (Ti-6Al-4V, butt joint)

No Phase Phase S22 Transformation Transformation (MPa)S22 800 700 600 500 400 PWHT No 300 200 100 0 -100

PWHT

. Figure 148 Y stress distribution in the welding direction viewed from the top surface and close up weld transverse cross-section (20 mm wide) with or without post weld heat treatment (PWHT) as well as either taking into account the effect of phase transformation (Ti-6Al-4V, butt joint)

244 Figure 146 and Figure 147 show that introducing the phase transformation effect to the numerical weld model only had a small influence on the resultant magnitude and distribution of residual stresses around the weld at the end of the welding process but some difference was observed with sheet distortions. The peak magnitude of y stress was found to be 50 MPa greater in the phase transformation model with a steeper stress gradient across the weld width with more variations in and near the FZ and the HAZ than in the model without phase transformation which remained almost constant in and near the FZ, while maintaining almost the same trend in RS distribution. Similarly, the calculated x and z stress magnitudes were again larger with phase transformation but only by a very small amount.

While both models with or without phase transformations over predicted the angular and cambering out of plane distortions when compared to the experimental measurements obtained using CMM, the simulation results from the model with phase transformation effect showed greater differences from the CMM results, than those from the model without phase transformation by around few millimetres. The distribution of out of plane displacements was therefore, found to be more sensitive than the residual stress distribution to the effect phase transformations. The reason for such negligible difference between the simulated welding residual stresses from both conditions, as mentioned above, was most likely due to well matched dimensions of α and β unit cells, meaning that there was no reason for any alterations of the initial crystallographic orientation during nucleation and dissolution of new phases. Consequently, the level of internal stresses due to phase transformation remained low unlike other materials which exhibit greater differences in the specific volumes between phases and therefore, no significant changes in the welding induced residual stresses were caused by phase transformation in Ti-6Al-4V. In fact, transformation induced volumes changes in titanium alloys are typically one order of magnitude smaller than those of ferrous alloys [423].

Figure 146 and Figure 148 show a reduced peak tensile y stress in the welding direction after applying PWHT from around 750 MPa before to 450 MPa after, so almost 40% reduction was achieved with PWHT. The magnitude of y stress after PWHT was similar to the value of its yield strength at 600°C at a very slow strain rate of 0.001 s-1 as a result of slow and uniform heating and cooling during PWHT. Even though the magnitude of the tensile y residual stress became lower, its extent became wider after PWHT compared to the as welded stress state. Almost no change occurred outside the weld region and constantly remained close to stress free state. The x and z stresses already had negligible magnitudes in the as welded condition but still they were further reduced after PWHT. However, the contributions from these two components when compared to the y stress on the final residual stress states were insignificant. The differences between the stress values from the models with and without phase transformation were even smaller after PWHT so the effect of phase transformation

245 became less after PWHT. Figure 146 also illustrates the evolution of out of plane displacements after PWHT. Under carefully controlled heat treatment at the correct heating and cooling rate, treatment temperature and soaking time, reduction in residual stresses was achieved by release of locked-in stresses. However, PWHT also resulted in greater distortions, with a small increase in the maximum angular distortion by around 0.5 mm both with or without phase transformation. A slightly greater increase in cambering distortion of around 2.0 mm after PWHT was simulated with phase transformation compared to around 1.5 mm without. While there was a relatively small effect of phase transformation on residual stresses, there were noticeable difference in welding distortions. It would therefore be necessary to further conduction experimental measurements on PWHT welded specimens to validate the numerical simulation results and to examine the influence of phase transformation on residual stresses and distortions after PWHT, as there were only as welded specimens accessible at the time of this investigation.

6.7.4 Mechanical Stress Relieving (MSR) Treatment It is often difficult to apply thermal treatments to very large welded structures to stress relieve high levels of inherent residual stresses in order to ensure the reliability of these structures during service. In addition, the thermal treatments can also result in undesirable effects such as loss of strength and increased distortions which may not be acceptable. In particular, buckling distortion can be a serious problem when welding thin sheet metals of thickness less than 4 mm due to their low critical buckling stress and also applying post weld distortion control methods generates non value added manufacturing costs and time [448]. In the past, mitigation methods have been developed by many researchers in order to reduce residual stresses and distortions using numerous methods, including for example, low stress no distortion (LSND) [449] technique, transient or post weld mechanical or thermal tensioning, trailing heat sink and optimised welding sequences.

Mechanical stress relieving (MSR) is an alternative method to PWHT for reducing residual stresses and distortions, by means of plastic deformation applied during as well as after welding [450]. The advantage of MSR is that it can be used as an active in-process method during welding without having to perform additional post weld machining operation so it is more convenient and economical than PWHT in large structures due to the fact that PWHT requires large heating facilities. MSR by mechanical tensioning involves applying uniform global external pre-tensile loading in the welding direction before and during welding to the entire workpiece including the vicinity of weld and subsequent removal after welding [450,451]. Elastic pre-tension changes the transient stress distribution during welding. It counteracts the compressive stresses during heating which has the effect of reducing the residual plastic compressive strains in the weld. On the other hand, it causes plastic deformation in the highly

246 tensile weld region during cooling as the weld can no longer sustain the superposed external tensile load any further and as a result, reduces the peak value of y residual stresses in the weld zone and produces a more uniform distribution after unloading [452].

Relatively little work has been conducted in the past to examine the underlying mechanism of mechanical tensioning and verify its effectiveness [453]. Qi and Zhang [454] analysed the effect of elastic pre-tensioning on the critical buckling stress in tungsten inert gas (TIG) welded thin Ti-6Al-4V sheets and determined the critical y compressive stress that would lead to buckling deformation based on tensile stress and plastic zone width from FEM. Elastic tensioning caused redistribution of incompatible strains in the weld and lowered the maximum compressive stresses below that of the critical buckling stress. Xu et al. [455] investigated mechanical stress relieving of submerged arc welded (SMAW) A105 forging steel without post- weld heat treatment by simulating a 2D axisymmetric FE model and blind hole drilling to measure the residual stresses on the outer surface before and after MSR. Richards et al. [456], Altenkirch et al. [451] and Staron et al. [457] examined the effectiveness of transient mechanical tensioning for controlling residual stresses in aluminium alloy friction stir welds by simulating and measuring the residual stress state in welds that had been tensioned to different loads during welding using synchrotron X-ray and neutron diffraction techniques. It was determined that low levels of tensioning reduced the magnitude of tensile peak stresses as well as distortions, and even led to compressive stress state in the weld at high tensioning loads, typically at levels greater than 35-50% of the room temperature yield strength of the base metal.

S22 (MPa) σ=5% YS σ=30% YS 800 700 600 500 σ=10% YS σ=50% YS 400 300 200 100 σ=20% YS σ=90% YS 0 -100

σ=5% YS σ=10% YS σ=20% YS σ=30% YS σ=50% YS σ=90% YS Figure 149 Y stress distirbution after welding when pre-loaded to 5, 10, 20, 30, 50 and 90% of the room temperature yield stress of the base metal by mechanical tensioning, view from the weld transverse section and top surface (Ti-6Al-4V, butt joint)

247 800 S22 (0% YS) 700 S22 (5% YS) S22 (10% YS) S22 (20% YS) 600 S22 (30% YS) S22 (50% YS) S22 (90% YS) 500

400

300 Stress (MPa) 200

100

-100 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Figure 150 Residual stress distributions in the y direction as a function of the applied tensioning level, pre- loaded to 5, 10, 20 ,30, 50 and 90% of the room temperature yield sterngth of the base metal (Ti-6Al-4V, butt joint)

In this investigation, to simulate mechanical tensioning, mechanical boundary conditions were applied to fix the workpiece on one end by limiting degrees of freedom in the desired directions to allow movement when the global tensile loads were applied uniformly to the other end of the workpiece in the welding direction. In this case, other clamping fixtures were not modelled. The simulated loads were 5, 10, 20, 30, 50 and 90% of the room temperature yield strength of the base metal. The load was applied before the start of welding and then removed after cooling down to room temperature.

Figure 150 show the variations of the residual stress distributions with mechanical tensioning at different load levels. Detailed stress distributions in the x and z directions can be found in Appendix B. It was obvious that the stresses in the x and z directions were typically much smaller than the y stresses and found to be largely insensitive to the tensioning level, so almost no change was identified. On the other hand, the effect of elastic pre-tension on the relaxation behaviour of the y residual stresses was largely determined by the magnitude of the applied mechanical load. The magnitude of y residual stresses decreased uniformly across the weld with almost no alteration of the stress distribution with increasing pre-tension load. The peak residual stress dropped by almost half at 30% YS and to around ¼ at 50% YS. When the load reached more than 50% YS, as shown by the stress contour at 90% YS, the tensile residual stress near the weld became from tensile to mostly compressive. Therefore, mechanical tensioning during welding clearly had a positive effect on reducing the welding induced y residual stresses while not affecting the other x and z stresses which were negligible anyways.

248 In addition, both cambering and angular distortions were reduced significantly by pre- tensioning as shown in Figure 151. Even applying very small loads of 5% and 10% YS led to considerable reductions in out of plane displacements. At a load level equal to or greater than 20% YS resulted in almost no cambering distortion and no further reduction in angular distortion even at higher loads. This means that mechanical tension can be an effective method for reducing both residual stresses and distortions even at relatively small loads of around 30% YS. It would be necessary to further validate the simulation results by conducting welding experiments using mechanical tensioning at different loads and measure the residual stress distributions using various measurement techniques such as X-ray and neutron diffraction techniques as discussed previously. a) b) 5.0 3.0 0% YS 0% YS 4.5 5% YS 5% YS 10% YS 10% YS 20% YS 2.5 20% YS 4.0 30% YS 30% YS 50% YS 50% YS 90% YS 3.5 90% YS 2.0 3.0

2.5 1.5

2.0 plane displacementplane (mm)

plane displacementplane (mm) 1.0 -

- 1.5

- Out Out 1.0 0.5 0.5

0.0 0.0 -250 -200 -150 -100 -50 0 50 100 150 200 250 -200 -150 -100 -50 0 50 100 150 200

Distance along specimen length (mm) Distance from sepcimen centre (mm) Out displacement of (mm) Out plane Y distance(mm) displacement of (mm) Out plane Distance from weld centre (mm)

Figure 151 a) Cambering and b) angular out of plane displacements as a function of the applied tensioning level, pre-loaded to 5-90% of the room temperature yield sterngth of the base metal (Ti-6Al-4V, butt joint) 6.8 Conclusions Numerical simulation results of mechanical analysis showed good agreement in general with the experimental results in terms of the residual stress distributions and magnitudes. The welded specimens were characterised by a more pronounced dominance of y residual stresses over x and z residual stresses. The y residual stresses in all the cases were very high but not as high as the yield strength of the material at room temperature. They were largely tensile in nature only within the FZ or HAZ and tended to be weakly compressive in the rest of the specimen.

Mechanical boundary condition had relatively small influence on residual stresses in thin sheets of butt welded specimens, whereas, greater restraints led to higher residual stresses and lower restraints led to lower residual stresses in T-joint specimens. Welding distortion was clearly influenced by the clamping conditions, where the minimum distortion was induced by

249 clamping close to the weld without affecting the residual stresses significantly in the butt welded specimens but increased them considerably in the T-joint specimens.

Welding parameters were found to influence the residual stress fields. Welding with filler metal resulted in a wider tensile region in the u direction, higher peak tensile and compressive stress magnitudes around the weld and lower u stresses in the softened weld metal. Similarly, greater out of plane displacements were detected when welding with filler metal. Heat input per unit volume was also found to be an influential parameter for the control of residual stress levels.

Phase transformations only had a small influence on the magnitude and distribution of residual stresses because the level of internal stresses due to phase transformation remained low unlike other materials which exhibit greater differences in the specific volumes between phases. However, larger component distortions were observed with phase transformations.

Post weld heat treatment induced diffusional phase transformations via decomposition of martensite into α. It also decreased the magnitude of u stresses to the yield strength of Ti-6Al- 4V at the treatment temperature by releasing the locked-in stresses. Mechanical stress relieving was also studied as an alternative method to PWHT for reducing residual stresses and distortions, by means of plastic deformation applied during as well as after welding. The x and z stresses were found to be largely insensitive to different elastic tension loads, whereas, the peak y stresses dropped by almost half at a load equal to 30% and to around a quarter at a load equal to 50% of the yield strength of the material at room temperature. When the load reached more than 50% YS, the stresses became compressive.

250 7 CONCLUSIONS AND FUTURE WORK

7.1 Conclusions Metallographic weld evaluation was performed at both macro and micro levels on fibre laser welded specimens made out of AA 2024-T3 and Ti-6Al4V. The influence of various processing parameters including power density, laser power, welding speed, focal position, filler metal feed rate and shielding gas composition on the weld quality, weld bead geometry, and weld microstructure was investigated. It was possible to weld AA 2024-T3 and Ti-6Al-4V over a wide range of processing parameters, with characteristics and qualities meeting strict requirements for advanced aerospace applications.

AA 2024-T3 consisted of fine equiaxed and columnar dendrites in the fusion zone. The heat affected zone consisted of a partially melted zone close to the weld and an overaged zone further away. The base metal had a triple phase eutectic structure containing copper and magnesium as the main alloying elements. Ti-6Al-4V consisted of a fine martensitic microstructure in the fusion zone and mostly martensitic and acicular α structure in the adjacent heat affected zone, and primary α, intergranular β and martensite close to the base metal. The base metal consisted of primary α and intergranular β.

Heat input supplied to the workpiece was mainly controlled by laser power and welding speed, which in turn, influenced the weld microstructure by modifying the peak temperatures experienced and the heating and cooling rates. For AA 2024-T3, the dissolution of strengthening phases in the FZ was more pronounced at higher temperatures. Faster welding speed or steeper thermal gradients led to more elongated weld pool which promoted formation of columnar dendritic structures, whereas, lower welding speeds led to larger dendrite cells in the FZ, wider FZ width and larger grains in the HAZ. Equiaxed dendrites were formed at the weld centreline at lower welding speeds. For Ti-6Al-4V, decreasing the laser power or increasing the welding speed resulted in finer martensite and prior β grains in the FZ, whereas, increasing the laser power or decreasing the welding speed resulted in acicular martensitic structure and larger prior β grain size, as well as the formation of diffusional transformation constituents such as acicular α and grain boundary α. No significant differences in weld microstructure were observed over the range of focal positions examined.

Both the weld top and bottom width increased with increasing laser power, focal position and decreasing welding speed. The hot cracking sensitivity of AA 2024-T3 was reduced by controlling the weld shape as it dictated the solidification pattern, with a wider weld pool reducing the risk of solidification cracking, shrinkage strain during solidification and the weld metal composition. Incomplete penetration or narrow root width were the main problems at

251 low laser powers and fast welding speeds, whereas, undercut was the main defect at high laser powers. The weld quality was on average good for all welding speeds as long as it was not too fast. Spatter was observed at the bottom surface at very fast welding speeds. Reinforcement increased with increasing laser power or welding speed but was not critical enough to affect the weld quality. The root width increased more at negative focal positions than at the equivalent positive focal positions due to an enhanced interaction between laser and the material being welded with negative defocusing.

Welding with filler metal reduced the crack sensitivity of AA 2024-T3 but it was also important to optimise the filler metal feed rate to avoid the formation of welding defects and keyhole instability. For AA 2024-T3, the weld quality was improved using helium because the plume effect was reduced due to the higher ionisation potential of helium which minimised porosity formation.

Mechanical tests were also conducted to determine the thermo-mechanical properties AA 2024-T3 and Ti-6Al-4V welds. Micro-indentation hardness testing showed that for AA 2024- T3 the hardness in the FZ was the lowest, higher in the HAZ and the highest in the BM. Softening in the FZ and over-aging in the HAZ were caused by the dissolution or coarsening of the strengthening precipitates during welding where very high temperatures as well as rapid heating and cooling rates were experienced. For Ti-6Al-4V, the FZ exhibited the largest hardness values due to the formation of fine needle like martensitic microstructure at fast cooling rates. The hardness in the HAZ gradually dropped with increasing distance from the weld centre due to a progressive drop in the martensite content. Increasing the heat input resulted in lower hardness in the FZ. For higher heat inputs, longer exposure at high temperatures and slower cooling rate led to the formation of a plate like martensitic phase instead of a needle like one, as well as diffusional transformation constituents such as acicular α and grain boundary α. It also caused the prior β grain size to increase. The global and local strain response of welded specimens were determined using the DIC. AA 2024-T3 welds were under-matched with the maximum joint efficiency of around 85%. The weaker strength of the FZ deteriorated the plastic straining capacity of the welded specimens due to increased stress concentration and constraints in the FZ which confined plasticity development within the weld. Ti-6Al-4V welds were over-matched with the minimum local plastic deformation in the FZ and the global tensile properties of the welded specimens were predominantly determined by the properties of the BM with joint efficiencies were close to 100%.

Finite element models were developed to predict welding induced residual stresses and distortion of test plates and T-joints, and validated using experimental database including weld pool geometry and temperature fields. X-ray and neutron diffraction measurements were performed to experimentally determine residual stresses in the welded specimens. The

252 specimens investigated were of laboratory scale and far smaller than the size of real aircraft components in manufacturing industry. However, once validated they may serve as foundation for larger aircraft components which satisfy industrial requirements. The specimens were characterised by a more pronounced dominance of y residual stresses over x and z residual stresses. The observed deviations in the simulated peak y stress levels from neutron diffraction measurements were mainly caused by the stress free lattice parameters used when processing the elastic strains because of chemical compositional and microstructural variations in and around the weld. The use of volume averaged local values or plane stress assumption led to better agreement with the numerical results.

The measured residual stresses were dependent on the crystallographic hkl plane from which they were obtained from and the cause of such difference was found to be due to the presence of microscopic stresses. However, in some cases by fitting the entire diffraction spectrum with Rietveld refinements instead of conducting a single peak analysis, it was possible to use the macroscopic elastic constants and the results were in good agreement with the numerical results. In the case of Ti-6Al-4V, the reflections were weak and only few times larger than the background due to its highly incoherent cross-section and texture, thus making neutron diffraction measurements difficult. The best agreement with the numerical results was found with the {201} peak.

Mechanical boundary conditions had relatively small influence on residual stresses in thin sheets of butt welded specimens, whereas, greater restraints led to higher residual stresses and lower restraints led to lower residual stresses in T-joint specimens. In contrast, welding distortion was clearly influenced by them. The minimum distortion was achieved by restraining close to the weld region without affecting the residual stresses significantly in the butt welded specimens, whereas, they increased considerably in the T-joint specimens. Heat input per unit volume was also found to be an influential parameter for the control of residual stress levels. With increasing heat input or the amount of heat energy supplied during welding, the stress level increased.

Non-isothermal diffusional and diffusionless phase transformations in Ti-6Al-4V were modelled and their influence on the residual stresses and distortions was examined. It only had a small influence on the magnitude and distribution of residual stresses because the level of internal stresses due to phase transformation remained low unlike other materials which exhibit greater differences in the specific volumes between phases. However, larger component distortions were observed with phase transformations. Post weld heat treatment and mechanical stress relieving processes were modelled for reducing residual stresses and distortions.

253 7.2 Future Work This thesis only focused upon selective aspects of fibre laser welding AA 2024-T3 and Ti-6Al- 4V so there still remains some room for improvement. Some aspects which require further investigation for future work which have come to light during this work are described below.

Rapid heating and cooling rate observed in laser welding mean that a very narrow HAZ is produced. This makes the examination of microstructure and mechanical properties of the HAZ very difficult. Simulating the welding thermal cycle allows us to obtain a large volume of the same microstructure as at a certain position in the HAZ. It is more convenient and effective to analyse specimens with a microstructure produced by HAZ simulation rather than characterising the HAZ by examining the actual welds. Unfortunately, only preliminary experiments were conducted to try simulating the welding thermal cycle using the Gleeble simulator due to limitations of procedures in experiments. Water quenching was used to achieve the required high cooling rate, but it also generated a steep temperature gradient across the specimen and resulted in a non-uniform microstructure and mechanical properties. One possible method of solving with this problem would be to use a special isothermal quenching (ISO-Q) technique, which would involve drilling a hole at each end of a round specimen and then water quenching at both ends of the reduced section instead of direction quenching. In this way, an isothermal plane could be maintained during cooling in the mid span since heat losses would occur along the specimen’s axis.

Further investigation may be performed to examine the fracture toughness and fatigue resistance including fatigue crack growth rate (da/dN) and fatigue life (S-N) of fibre laser welded AA 2024-T3 and Ti-6Al-4V. Unstiffened fatigue crack specimens and stiffened panel fatigue specimens with 3-4 stringers of around 1 m in panel length and width and 200 mm stringer pitch could be used for example, with a long crack introduced on the weld line under static and fatigue loading to investigate the influence of joint properties on fatigue resistance.. Measurement of fatigue crack growth rates would be performed at constant load amplitude at 10~20 Hz and different R values, with welds loaded parallel to their y axis and cracks growing across them. It would be necessary to determine plastic strain distribution in and around welded joints and also at the crack tip to understand fracture process. It would involve performing CTOD tests on standard compact tension (CT) specimens extracted from welded plate and notched in the FZ and HAZ to determine the local fracture toughness properties.

Numerical simulations could be done to predict fatigue crack growth and fracture toughness behaviour of the welded joints subject to Mode I loading condition, and also identify any influence of residual stress on these properties. It is likely that the presence of residual stresses can significantly affect the fatigue behaviour of welded structures. For example,

254 tensile residual stresses are known to have a negative influence on the fatigue life as they increase the rate of fatigue crack growth. Accurate prediction of welding induced residual stresses as demonstrated in this thesis are therefore, essential for structural integrity and fatigue life assessment of the welded part.

255 REFERENCES [1] Mirza FA, Chen DL. Fatigue of Magnesium Alloys. In: Zhang S, Zhao D (eds.) Aerospace Materials Handbook. Boca Ration, FL: CRC Press; 2012. p. 647–698.

[2] Ratchev P, Verlinden B, De Smet P, Van Houtte P. Precipitation hardening of an Al– 4.2wt% Mg–0.6wt% Cu alloy. Acta Materialia. 1998;46(10): 3523–3533.

[3] Ding RG, Ojo OA, Chaturvedi MC. Fusion zone microstructure of laser beam welded directionally solidified Ni3Al-base alloy IC6. Scripta Materialia. 2006;54: 859–864.

[4] Katayama S, Kawahito Y, Mizutani M. Latest Progress in Performance and Understanding of Laser Welding. Physics Procedia. 2012;39: 8–16.

[5] Dittrich D, Standfuss J, Liebscher J, Brenner B, Beyer E. Laser beam welding of hard to weld Al alloys for a regional aircraft fuselage design- First results. Physics Procedia. 2011;12: 113–122.

[6] Mendez PF. Welding Processes used in the Aeronautic Industry Friction Welding (FRW). New Trends for the Manufacturing in the Aeronautic Industry. 2000. p. 21–38.

[7] Rendigs KH. Aluminium Structures Used in Aerospace - Status and Prospects-. Materials Science Forum. 1997;242: 11–24.

[8] Koçak M, Seib E, Motarjemi A. Treatments of Structural Welds using FITNET Fitness- for-Service Assessment Procedure. Welding in the World. 2013;51(5-6): 106–118.

[9] Seib E, Koçak M. Fracture Analysis of Strength Undermatched Welds of Thin-Walled Aluminium Structures Using Fitnet Procedure. Welding in the World. 2013;49(11-12): 58–69.

[10] Koçak M. Structural Integrity of Welded Structures : Process - Property - Performance (3P) Relationship. 63 Annual Assembly & International Conference of the International Institute of Welding. 2010. p. 3–19.

[11] Dos Santos J, Ç am G, Torster F, Insfran A, Riekehr S, Ventzke V, et al. Properties of power beam welded steels, al-and ti-alloys: Significance of strength mismatch. Welding in the World. 2000;44(6): 42–64.

[12] Cao X, Wallace W, Immarigeon J-PP, Poon C. Research and Progress in Laser Welding of Wrought Aluminum Alloys. II. Metallurgical Microstructures, Defects, and Mechanical Properties. Materials and Manufacturing Processes. 2003;18(1): 23–49.

[13] Farahmand B, Saff C, Xie D, Abdi F. Estimation of Fatigue and Fracture Allowables For Metallic Materials Under Cyclic Loading. 48th AIAA/ASME/ASCE/AHS/ASC

256 Structures, Structural Dynamics, and Materials Conference. Reston, Virigina: American Institute of Aeronautics and Astronautics; 2007. p. 1–12.

[14] De A, Maiti SK, Walsh CA, Bhadeshia HKDH. Finite element simulation of laser spot welding. Science and Technology of Welding and Joining. 2003;8(5): 377–384.

[15] Miyamoto I, Park S, Ooie T. Ultra-fine Keyhole Welding with Single-mode Fiber Laser. Proceedings of the Second International WLT-Conference on Lasers in Manufacturing. Munich: Laser Institute of America; 2003. p. 225.

[16] Quintino L, Costa A, Miranda R, Yapp D, Kumar V, Kong CJ. Welding with high power fiber lasers – A preliminary study. Materials & Design. 2007;28(4): 1231–1237.

[17] Elmer JW. Introduction to High Energy Density Electron and Laser Beam Welding. Welding Fundamentals and Processes. Vol 6A, ASM Handbook. ASM International; 2011. p. 507–513.

[18] ISO 13919-1 Welding — Electron and laser-beam welded joints — Guidance on quality levels for imperfections — Part 1: Steel.

[19] ISO 13919-2 Welding — Electron and laser-beam welded joints — Guidance on quality levels for imperfections — Part 2: Aluminium and its weldable alloys.

[20] AWS D17.1 Specification for fusion welding for aerospace applications.

[21] Chen YC, Feng JC, Liu HJ. Precipitate evolution in friction stir welding of 2219-T6 aluminum alloys. Materials Characterization. 2009;60(6): 476–481.

[22] Wang SC, Starink MJ. Two types of S phase precipitates in Al–Cu–Mg alloys. Acta Materialia. 2007;55(3): 933–941.

[23] Polmear I, St. John D. Wrought Aluminium Alloys. Light Alloys: From Traditional Alloys to Nanocrystals. 4th ed. Oxford: Butterworth-Heinemann; 2005. p. 134–171.

[24] Bartl J, Baranek M. Emissivity of aluminium and its importance for radiometric measurement. Measurement of Physical Quantities. 2004;4(3): 31–36.

[25] Kim JS, Watanabe T, Yoshida Y. Effect of the beam-defocusing characteristics on porosity formation in laser welding. Journal of Materials Science Letters. 1995;14(22): 1624–1626.

[26] Alfieri V, Caiazzo F, Cardaropoli F, Sergi V. Issues in Disk-laser Key-hole Welding of 2024 Aluminum Alloy. Frontiers in Aerospace Engineering. 2012;1(1): 36–44.

257 [27] Chaturvedi MC. Laser welding of metals for aerospace and other applications. In: Chaturvedi MC (ed.) Welding and joining of aerospace materials. 1st ed. Cambridge: Woodhead Publishing Limited; 2012. p. 75–108.

[28] Kou S. Weld Metal Solidification Cracking. In: Kou S (ed.) Welding Metallurgy. 2nd ed. Hoboken: John Wiley & Sons, Inc.; 2003. p. 263–300.

[29] Koçak M, dos Santos J, Riekehr S. Trends in laser beam welding technology and fracture assessment of weld joints. Science and Technology of Welding and Joining. 2001;6(6): 347–350.

[30] Anderson T. How to Avoid Cracking When Welding Aluminum Alloys. AWS Weld Cracking Conference. Denver, CO; 2014.

[31] Cao X, Wallace W, Poon C, Immarigeon JP. Research and Progress in Laser Welding of Wrought Aluminum Alloys. I. Laser Welding Processes. Materials and Manufacturing Processes. 2003;18(1): 1–22.

[32] Gittos MF. Welding Al-Mg-Si alloys. Welding Institute Research Bulletin. 1986;27: 231–235.

[33] Cicalǎ E, Duffet G, Andrzejewski H, Grevey D, Ignat S, Cicală E, et al. Hot cracking in Al-Mg-Si alloy laser welding - Operating parameters and their effects. Materials Science and Engineering A. 2005;395(1-2): 1–9.

[34] Yamaoka H. Microstructural control of laser‐ welded aluminium alloys. Welding International. 2001;15(11): 845–850.

[35] Zhao H, White DR, DebRoy T. Current issues and problems in laser welding of automotive aluminium alloys. International Materials Reviews. 1999;44(6): 238–266.

[36] Kanazawa H. Welding performance of high power YAG lasers with aluminium alloys. Welding International. 1998;12(5): 341–346.

[37] Rudy JF, Rupert EJ. Effects of porosity on mechanical properties of aluminium welds. Welding Journal. 1970;49(7): 322s – 337s.

[38] Mazur M. Porosity in aluminium welds. Welding international. 1992;6(12): 929–931.

[39] Kroos J, Gratzke U, Simon G. Towards a self-consistent model of the keyhole in penetration laser beam welding. Journal of physics D: Applied physics. 1993;26(3): 474.

[40] Anyalebechi PN. Analysis of the effects of alloying elements on hydrogen solubility in liquid aluminum alloys. Scripta Metallurgica et Materialia. 1995;33(8): 1209–1216.

258 [41] Mathers G. Welding design. In: Mathers G (ed.) The Welding of Aluminium and its Alloys. 1st ed. Boca Raton, FL: Woodhead Publishing Limited; 2002. p. 69–96.

[42] Zhao H, Debroy T. Weld Metal Composition Change during Conduction Mode Laser Welding of Aluminum Alloy 5182. Metallurgical and Materials Transactions B. 2001;32(1): 163–172.

[43] Ramasamy S, Albright CE. CO2 and Nd:YAG laser beam welding of 6111-T4 aluminum alloy for automotive applications. Journal of Laser Applications. 2000;12(3): 101.

[44] Ramasamy S, Albright CE. CO2 and Nd:YAG laser beam welding of 5754–O aluminium alloy for automotive applications. Science and Technology of Welding and Joining. 2001;6(3): 182–190.

[45] Cieslak MJ, Fuerschbach PW. On the weldability, composition, and hardness of pulsed and continuous Nd:YAG laser welds in aluminum alloys 6061,5456, and 5086. Metallurgical Transactions B. 1988;19(2): 319–329.

[46] Zhao H, Martukanitz RP, Debroy T. Porosity, Underfill and Magnesium Loss during Continuous Wave Nd : YAG Laser Welding of Thin Plates of Aluminum Alloys 5182 and 5754. Welding Journal. 1999;78(6): 207–216.

[47] Mouritz AP. Aluminium alloys for aircraft structures. Introduction to Aerospace Materials. 1st ed. Cambridge: Woodhead Publishing; 2012. p. 173–201.

[48] Liu J, Kulak M. A New Paradigm in the Design of Aluminum Alloys for Aerospace Applications. Materials Science Forum. 2000;331-337: 127–142.

[49] Liu J. Advanced Aluminium and Hybrid Aerostructures for Future Aircraft. Materials Science Forum. 2006;519-521: 1233–1238.

[50] Palm F. Residual strength in integral (welded) Al fuselage structures. International Conference on Trends in Welding Research. 2005.

[51] Ion JC. Laser beam welding of wrought aluminium alloys. Science and Technology of Welding and Joining. IOM Communications for the Institute of Materials, 2000;5(5): 265–276.

[52] Mouritz AP. Materials and material requirements for aerospace structures and engines. Introduction to Aerospace Materials. 1st ed. Cambridge: Woodhead Publishing; 2012. p. 39–56.

259 [53] Bray GH, Bucci . J, Kulak M, Warren CJ, Grandt AF, Golden PJ, et al. Benefits of improved fuselage skin sheet alloy 2524-T3 in multisite damage scenarios. Light metal age. Fellom, 56(11-12).

[54] Williams JC, Starke EA. Progress in structural materials for aerospace systems. Acta Materialia. 2003;51(19): 5775–5799.

[55] Ion JC. Keyhole welding. In: Ion JC (ed.) Laser Processing of Engineering Materials: Principles, Procedure and Industrial Application. 1st ed. Oxford: Butterworth- Heinemann; 2005. p. 395–448.

[56] Mendez PF, Eagar TW. New trends in welding in the aeronautic industry. Proceedings of the conference ‘New Trends for the Manufacturing in the Aeronautic Industry’. San Sebastián: Hegan and Inasmet; 2000. p. 21–38.

[57] Dearden G, Schleyer GK, Watkins KG, Okon P, Simmons MC, Okon P, et al. Blast and impact resistance studies of laser welded and riveted panel structures. Proceedings of the 21st International congress on applications of lasers and electro- optics. Scottsdale, AZ: Laser Institute of America; 2002. p. 1–11.

[58] Kocik R, Kaschel S, Kreimeyer M, Vollertsen F, Gmbh AD, Technology M. Development of a new joining technology for hybrid metal aircraft structures. 23rd International Congress on Applications of Lasers & Electro-Optics. San-Francisco, CA: Laser Institute of America; 2004.

[59] Zink W. Integral solutions for fuselage shells. In: Peters M, Kaysser WA (eds.) Proceedings of the 19th European Conference on Advanced Aerospace Materials- Materials for Aerospace Applications. Munich: DGLR-Bericht; 2000. p. 25–35.

[60] Kokkonen J, Ion JC. CO2 Laser Welding of the Aluminium Aerospace Alloy AA2024. In: Veli K, Ion J (eds.) Proceedings of the 7th Nordic Conference on Laser Advanced Materials Processing. Lappeenranta: Acta Universitatis Lappeenrantaensis; 1999. p. 418–429.

[61] Neye G, Heider P. Laser beam welding of modern Al-alloys for the aircraft industry. Proceedings of the 5th European Conference on the Laser Treatment of Materials. Bremen: DVS Berichte; 1994. p. 108–117.

[62] Liu J, Watanabe I, Yoshida K, Atsuta M. Joint strength of laser-welded titanium. Dental Materials. 2002;18(2): 143–148.

260 [63] Chaturvedi MC. New welding techniques for aerospace engineering. Welding and joining of aerospace materials. 1st ed. Cambridge: Woodhead Publishing Limited; 2012. p. 3–9.

[64] Norris G, Wagner M. High-Tech Advantage. In: Gansen S, Hitch L (eds.) Airbus A380: Superjumbo of the 21st Century. 1st ed. St. Paul, MN: Zenith Press; 2005. p. 61–70.

[65] Palm F. Can welded fuselage structures fulfil future A/C damage tolerance requirements. In: Benedictus R (ed.) First International Conference on Damage Tolerance of Aircraft Structure. Delft: Aerospace Materials and Structures Group; 2007.

[66] Abbaschian R, Hill RR. Selected Nonferrous Alloy Systems. Physical Metallurgy Principles. 4th ed. Stamford: Cengage Learning; 2009. p. 677–679.

[67] Mouritz AP. Titanium alloys for aerospace structures and engines. Introduction to Aerospace Materials. 1st ed. Cambridge: Woodhead Publishing; 2012. p. 202–223.

[68] Sieniawski J, Ziaja W, Kubiak K, Motyka M. Microstructure and Mechanical Properties of High Strength Two-Phase Titanium Alloys. In: Sieniawski J, Ziaja W (eds.) Titanium Alloys: Advances in Properties Control. Intech; 2013. p. 69–80.

[69] Boyer RR. An overview on the use of titanium in the aerospace industry. Materials Science and Engineering: A. 1996;213(1-2): 103–114.

[70] Manero J., Gil F., Planell J. Deformation mechanisms of Ti–6Al–4V alloy with a martensitic microstructure subjected to oligocyclic fatigue. Acta Materialia. 2000;48(13): 3353–3359.

[71] Abbaschian R, Hill RR. Titanium alloys. Physical Metallurgy Principles. 4th ed. Stamford: Cengage Learning; 2009. p. 322–327.

[72] Chen L, Shuili G, Yang J. Study on Full Penetration Stability of Light Alloys Sheet Laser Welding. In: Hinduja S, Li L (eds.) Proceedings of the 37th International MATADOR Conference. Manchester: University of Manchester; 2012. p. 315–318.

[73] Inoue H, Ogawa T. Weld cracking and solidification behavior of titanium alloys. Welding journal. 1995;74(1): 21s – 27s.

[74] Leyens C, Peters M. Titanium alloys for aerospace applications. In: Leyens C, Peters M (eds.) Titanium and Titanium Alloys. Fundamentals and Applications. Weinheim: WILEY-VCH Verlag GmbH & Co. KGaA; 2003. p. 333–351.

[75] Crane FAA, Charles JA, Furness J. Materials for airframes. Selection and Use of Engineering Materials. 3rd ed. Oxford: Butterworth-Heinemann; 1997. p. 227–255.

261 [76] Merlin PW. Design and development of the blackbird: Challenges and lessons learned. Proceedings of the 47th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition. Orlando, Fl: American Institute of Aeronautics and Astronautics; 2009. p. 5–8.

[77] Kannatey-Asibu E. Types of Lasers. Principles of Laser Materials Processing. 1st ed. New Jersey: John Wiley & Sons, Inc.; 2009. p. 135–182.

[78] Verhaeghe G, Hilton P. Battle of the Sources-Using a high-power Yb-fibre laser for welding steel and aluminium. Welding in the World-London-. 2005;49: 188.

[79] Verhaeghe G. The fibre laser - a newcomer for material welding and cutting. Welding Journal. 2005;84(8): 56–60.

[80] BS EN ISO 11146:1999 Lasers and laser related equipment - Test methods for laser beam parameters - Beam Widths, Divergence Angles and Beam Propagation Factor.

[81] Tsukamoto S, Steen WM, Mazumder J. Laser Welding. Laser Material Processing. 4th ed. London: Springer-Verlag London; 2010. p. 199–249.

[82] Schuöcker D. Welding with Lasers. High Power Lasers in Production Engineering. London: Imperial College Press; 1999. p. 337–370.

[83] Dowden J. Laser Keyhole Welding: The Vapour Phase. In: Dowden J (ed.) The Theory of Laser Materials Processing - Heat and Mass Transfer in Modern Technology. 1st ed. Bristol: Canopus Academic Publishing Limited; 2009. p. 104–105.

[84] Hügel H. New solid-state lasers and their application potentials. Optics and Lasers in Engineering. 2000;34(4-6): 213–229.

[85] Al-Kazzaz H, Medraj M, Cao X, Jahazi M. Nd:YAG laser welding of aerospace grade ZE41A magnesium alloy: Modeling and experimental investigations. Materials Chemistry and Physics. 2008;109: 61–76.

[86] Szymanski Z, Hoffman J, Kurzyna J. Plasma plume oscillations during welding of thin metal sheets with a CW CO 2 laser. Journal of Physics D: Applied Physics. 2001;34(2): 189–199.

[87] Mazmudar CP, Patel K. A Literature Review on Effect of Laser Welding Parameters on Mechanical Properties and Microstructure. International Journal for Scientific Research & Development|. 2013;1(10): 2102–2105.

[88] French PW, Li T, Clowes J, Perrie W, Sharp M, Liu D, et al. Ultrafast Short Pulse Laser Material Processing of Aerospace Materials. Proceedings of the Fourth International WLT-Conference on Lasers in Manufacturing. Munich; 2007. p. 1–6.

262 [89] Xiao R, Zhang X. Problems and issues in laser beam welding of aluminum–lithium alloys. Journal of Manufacturing Processes. 2014;16(2): 166–175.

[90] Salminen A, Piili H, Purtonen T. The characteristics of high power fibre laser welding. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science. 2010;224(5): 1019–1029.

[91] Maeda K, Kumagai M, Namba K. Application of diode laser to welding of aluminium alloys. Welding international. 2003;17(1): 860–863.

[92] Walsh CA, Bhadeshia HKDH, Lau A, Matthias B, Oesterlein R, Drechsel J. Characteristics of High-Power Diode-Laser Welds for Industrial Assembly. Journal of Laser Applications. 2003;15(2): 68–76.

[93] Sanders PG, Keske JS, Leong KH, Kornecki G. High power Nd:YAG and CO2 laser welding of magnesium. Journal of Laser Applications. 1999;11(2): 96.

[94] Quintino L, Miranda R, Dilthey U, Iordachescu D, Banasik M, Stano S. Laser Welding of Structural Aluminium. [Online] Moreira PMGP, da Silva LFM, de Castro PMST (eds.) Structural Connections for Lightweight Metallic Structures. Berlin, Heidelberg: Springer Berlin Heidelberg; 2012. 33-57 p.

[95] Cooper LJ, Wang P, Williams RB, Sahu JK, Clarkson WA, Scott AM, et al. High- power Yb-doped multicore ribbon fiber laser. Optics Letters. 2005;30(21): 2906.

[96] Steen WM. Background and General Applications. In: Steen WM (ed.) Laser Material Processing. 3rd ed. London: Springer-Verlag London; 2003. p. 11–58.

[97] Popov S. IPG Laser GmbH fibre lasers – driving material processing markets. Proceedings of AILU workshop on fibre lasers - future of laser material processing. Cranfield; 2006.

[98] Gambling WA. The rise and rise of optical fibers. IEEE Journal of Selected Topics in Quantum Electronics. IEEE, 2000;6(6): 1084–1093.

[99] Wagner F. Laser Beam Micro-Welding with Single Mode Fibre Lasers. Australasian Welding Journal. 2005;50: 14.

[100] Sorrentino S. Weldability and mechanical behaviour of induction assisted thick welds in high strength steel. International Scientific Colloquium: Modelling of Electromagnetic Processing. 2008. p. 201–208.

[101] Miranda R, Costa A, Quintino L, Yapp D, Iordachescu D. Characterization of fiber laser welds in X100 pipeline steel. Materials & Design. 2009;30(7): 2701–2707.

263 [102] Howse DS, Scudamore RJ, Booth GS. The evolution of Yb fibre laser/MAG hybrid processing for welding of pipelines. In: Chung JS, Hong SW, Koo J, Komai T, Koterayama W (eds.) The Proceedings of the 15th International Offshore and Polar Engineering Conference. Seoul: International Society of Offshore and Polar Engineers; 2005. p. ISOPE – I – 05–364.

[103] Thomy BYC, Seefeld T, Vollertsen F, Vietz E. Application of Fiber Lasers to Pipeline. Welding journal. 2006;85(7): 30–33.

[104] Karlsson J, Ilar T, Kaplan AFH. Knowledge platform approach for fiber laser welding of high strength. Proceedings of the 12th NOLAMP conference. Copenhagen: FORCE Technology; 2009. p. 1–15.

[105] Karlsson J, Kaplan A. Fibre laser welding for lightweight design. ICALEO 2009 - International Congress on Applications of Laser and Electro-Optics. 2009. p. 1548– 1557.

[106] Karlsson J, Kaplan AFH. Analysis of a fibre laser welding case study, utilising a matrix flow chart. Applied Surface Science. 2011;257(9): 4113–4122.

[107] Bergmann JP, Patschger a., Bastick a. Enhancing Process Efficiency due to high Focusing with high Brightness Lasers – Applicability and constraints. Physics Procedia. 2011;12: 66–74.

[108] Bastick A, Bergmann JP, Patschger A. High brilliancy — which effects do small foci have on secure and efficient welding? Physics Procedia. 2010;5: 147–156.

[109] Norman PM, Kaplan AFH, Karlsson J. Classification and generalization of data from a fibre-laser hybrid welding case. Physics Procedia. 2010;5: 69–76.

[110] Vollertsen F, GrÜ nenwald S, Rethmeier M, Gumenyuk A, Reisgen U, Olschok S. Welding Thick Steel Plates with Fibre Lasers and GMAW. Welding in the World. 2010;54(3-4): R62–R70.

[111] Kaplan AFH. Modelling the Primary Impact of an Yb:Fibre Laser Beam Profile on the Keyhole Front. Physics Procedia. 2011;12: 627–637.

[112] Zhenglong L, Caiwang T, Yanbin C, Zhongshao S. Microstructure and Mechanical Properties of Fiber Laser-Metal Active Gas Hybrid Weld of X80 Pipeline Steel. Journal of Pressure Vessel Technology. 2012;135(1): 011403.

[113] Grupp M, Klinker K, Cattaneo S. Welding of high thicknesses using a fibre optic laser up to 30 kW. Welding International. 2013;27(2): 109–112.

264 [114] Kawahito Y, Mizutani M, Katayama S. Investigation of high-power fibre laser welding phenomena of stainless steel. Transactions of JWRI. 2007;36(2): 11–16.

[115] Katayama S, Kawahito Y, Mizutani M. Elucidation of laser welding phenomena and factors affecting weld penetration and welding defects. Physics Procedia. 2010;5: 9– 17.

[116] Kawahito Y, Mizutani M, Katayama S. Elucidation of high-power ﬁbre laser welding phenomena of stainless steel and effect of factors on weld geometry. Journal of Physics D: Applied Physics. 2007;40(19): 5854–5859.

[117] Kawahito Y, Terajima T, Kimura H, Kuroda T, Nakata K, Katayama S, et al. High- power fiber laser welding and its application to metallic glass Zr55Al10Ni5Cu30. Materials Science and Engineering: B. 2008;148(1-3): 105–109.

[118] Salminen A, Lehtinen J, Harkko P. The effect of welding parameters on keyhole and melt pool behavior during laser welding with high power fibre laser. In: Guo W, Wang ZB, Li L, Liu Z, Luk’yanchuk B, Whitehead DJ (eds.) Proceedings of the 27th International Congress on Applications of Lasers and Electro-Optics. Temecula: Laser Institute of America; 2008. p. 703.

[119] Kaplan AFH, Wiklund G. Advanced welding analysis methods applied to heavy section welding with a 15 kW fiber laser. In: Sun Z (ed.) Proceedings of the International Conference on Welding of the IIW. Singapore: Singapore Welding Society; 2009. p. 295–300.

[120] Minhyo S, Kazuhiro N. Weld bead formation by a 10 kW class high power fiber laser on 16 mm thickness carbon steel plate. Transactions of JWRI. 2010;39(1): 33–38.

[121] Hartwig L, Ebert R, Kloetzer S, Weinhold S, Drechsel J, Peuckert F, et al. Material processing with a 3kW single mode fibre laser. Journal of Laser Micro Nanoengineering. 2010;5(2): 128–133.

[122] Eriksson I, Powell J, Kaplan AFH. Measurements of fluid flow on keyhole front during laser welding. Science and Technology of Welding & Joining. 2011;16(7): 636–641.

[123] Eriksson I, Powell J, Kaplan AFH. Melt flow measurement inside the keyhole during laser welding. In: Halmøy E, Fostervoll H, Bjørnbakk B (eds.) Proceedings of the 13th Conference on Laser Materials Processing in the Nordic Countries. Trondhelm: Norwegian University of Science and Technology; 2011.

265 [124] Cao X, Wanjara P, Huang J, Munro C, Nolting a. Hybrid fiber laser – Arc welding of thick section high strength low alloy steel. Materials & Design. Elsevier Ltd, 2011;32(6): 3399–3413.

[125] Zhang L, Lu JZ, Luo KY, Feng AX, Dai FZ, Zhong JS, et al. Residual stress, micro- hardness and tensile properties of ANSI 304 stainless steel thick sheet by fiber laser welding. Materials Science and Engineering: A. 2013;561: 136–144.

[126] Cui CY, Cui XG, Ren XD, Liu TT, Hu JD, Wang YM. Microstructure and microhardness of fiber laser butt welded joint of stainless steel plates. Materials & Design. 2013;49: 761–765.

[127] Coelho RS, Corpas M, Moreto JA, Jahn A, Standfuß J, Kaysser-Pyzalla A, et al. Induction-assisted laser beam welding of a thermomechanically rolled HSLA S500MC steel: A microstructure and residual stress assessment. Materials Science and Engineering: A. 2013;578: 125–133.

[128] Jasnau U, Sumpf A. Laser hybrid welding with high power fibre Laser-new chances for use of Laser technology. Nordic Welding Conference. 2006.

[129] Westerbaan D, Parkes D, Nayak SS, Chen DL, Biro E, Goodwin F, et al. Effects of concavity on tensile and fatigue properties in fibre laser welding of automotive steels. Science and Technology of Welding and Joining. 2014;19(1): 60–68.

[130] Meco S, Pardal G, Ganguly S, Williams S, McPherson N. Application of laser in seam welding of dissimilar steel to aluminium joints for thick structural components. Optics and Lasers in Engineering. 2015;67: 22–30.

[131] Sinha AK, Kim DY, Ceglarek D. Correlation analysis of the variation of weld seam and tensile strength in laser welding of galvanized steel. Optics and Lasers in Engineering. 2013;51(10): 1143–1152.

[132] Parkes D, Westerbaan D, Nayak SS, Zhou Y, Goodwin F, Bhole S, et al. Tensile properties of fiber laser welded joints of high strength low alloy and dual-phase steels at warm and low temperatures. Materials & Design. 2014;56: 193–199.

[133] Saha DC, Westerbaan D, Nayak SS, Biro E, Gerlich AP, Zhou Y. Microstructure- properties correlation in fiber laser welding of dual-phase and HSLA steels. Materials Science and Engineering: A. 2014;607: 445–453.

[134] Colegrove P, Williams SW, Ikeagu C, Thistlethwaite A. The Impact of Different Types of Welding Processes on the Residual Stress and Distortion in 4mm Thick Butt Welds

266 of Ship Plate. Trends in Welding Research, Proceedings of the 8th International Conference. 2009. p. 758–765.

[135] Chowdhury SM, Chen DL, Bhole SD, Powidajko E, Weckman DC, Zhou Y. Microstructure and Mechanical Properties of Fiber-Laser-Welded and Diode-Laser- Welded AZ31 Magnesium Alloy. Metallurgical and Materials Transactions A. 2011;42(7): 1974–1989.

[136] Chowdhury SH, Chen DL, Bhole SD, Powidajko E, Weckman DC, Zhou Y. Fiber Laser Welded AZ31 Magnesium Alloy: The Effect of Welding Speed on Microstructure and Mechanical Properties. Metallurgical and Materials Transactions A. 2012;43(6): 2133–2147.

[137] Yu L, Nakata K, Yamamoto N, Liao J. Texture and its effect on mechanical properties in fiber laser weld of a fine-grained Mg alloy. Materials Letters. 2009;63(11): 870–872.

[138] Kratky A, Schuöcker D, Liedl G. Processing with kW fibre lasers: advantages and limits. In: Vilar R, Conde O, Fajardo M, Silva LO, Pires M, Utkin A (eds.) XVII International Symposium on Gas Flow, Chemical Lasers, and High-Power Lasers. 2008. p. 71311X – 71311X – 12.

[139] Naffakh Moosavy H, Aboutalebi M-R, Seyedein SH, Goodarzi M, Khodabakhshi M, Mapelli C, et al. Modern fiber laser beam welding of the newly-designed precipitation- strengthened nickel-base superalloys. Optics & Laser Technology. 2014;57: 12–20.

[140] Ren W, Lu F, Yang R, Liu X, Li Z, Elmi Hosseini SR. A comparative study on fiber laser and CO2 laser welding of Inconel 617. Materials & Design. 2015;76: 207–214.

[141] Huntington CA, Eagar TW. Laser Welding of Aluminum and Aluminum Alloys. Welding Journal. 1983;62(4): 105–107.

[142] Leong KH, Sabo KR, Sanders PG, Spawr WJ. Laser Welding of Aluminum Alloys. Proceedings of Photonics West ’97. International Society for Optics and Photonics; 1997. p. 37–44.

[143] Leong KH, Sabo KR, Altshuller B, Wilkinson TL, Albright CE. Laser beam welding of 5182 aluminum alloy sheet . Journal of Laser Applications. 1999;11(3): 109–118.

[144] El-Batahgy A-M, Kutsuna M. Laser Beam Welding of AA5052, AA5083, and AA6061 Aluminum Alloys. Research Letters in Materials Science. 2009;2009: 1–9.

[145] Okon P, Dearden G, Watkins K, Sharp M, French P. Laser welding of aluminium alloy 5083. Proceedings of the 21st International Congress on Applications of Lasers and Electro-Optics. Scottsdale, AZ: Laser Institute of America; 2002. p. 1–9.

267 [146] Weston JP, Jones IA, Wallach ER. Laser welding of aluminium alloys using different laser sources. Proceedings of the 6th International Conference on Welding and Melting by Electron and Laser Beams. Toulon: International Institute of Welding; 1998.

[147] Sanchez-amaya JM, Boukha Z, Amaya-vazques MR, Botana FJ. Weldability of Aluminum Alloys with High-Power Diode Laser. Welding Research. 2012;91: 155– 161.

[148] Squillace A, Prisco U. Influence of filler material on micro- and macro-mechanical behaviour of laser-beam-welded T-joint for aerospace applications. Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications. 2009;223(3): 103–115.

[149] Shakeri HR, Buste a., Worswick MJ, Clarke J a., Feng F, Jain M, et al. Study of damage initiation and fracture in aluminum tailor welded blanks made via different welding techniques. Journal of Light Metals. 2002;2(2): 95–110.

[150] Choi K-D, Ahn Y-N, Kim C. Weld strength improvement for Al alloy by using laser weaving method. Journal of Laser Applications. 2010;22(3): 116.

[151] Fabrègue D, Deschamps A. Microstructural Study of Laser Welds Al6056-AS12 in Relation with Hot Tearing. Materials Science Forum. 2002;396-402: 1567–1572.

[152] Yang ZB, Tao W, Li LQ, Chen YB, Li FZ, Zhang YL. Double-sided laser beam welded T-joints for aluminum aircraft fuselage panels: Process, microstructure, and mechanical properties. Materials & Design. 2012;33: 652–658.

[153] Li M, Li Z, Zhao Y, Li H, Wang Y, Huang J. Influence of Welding Parameters on Weld Formation and Microstructure of Dual-Laser Beams Welded T-Joint of Aluminum Alloy. Advances in Materials Science and Engineering. 2011;2011: 1–6.

[154] Liu C, Northwood D., Bhole S. Tensile fracture behavior in CO2 laser beam welds of 7075-T6 aluminum alloy. Materials & Design. 2004;25(7): 573–577.

[155] Cui L, Li X, He D, Chen L, Gong S. Effect of Nd:YAG laser welding on microstructure and hardness of an Al–Li based alloy. Materials Characterization. 2012;71: 95–102.

[156] Bardin F, Mcbride R, Moore A, Morgan S, Williams S, Jones JDC, et al. Real time temperature measurement for process monitoring of laser conduction welding. Proceedings of the 23rd International Congress on Applications of Lasers and Electro- Optics. San Francisco, CA: Laser Institute of America; 2004. p. 1–24.

268 [157] Hu B, Richardson IM. Autogenous laser keyhole welding of aluminum alloy 2024. Journal of Laser Applications. 2005;17(2): 70–80.

[158] Alfieri V, Cardaropoli F, Caiazzo F, Sergi V. Porosity evolution in aluminum alloy 2024 bop and butt defocused welding by Yb-YAG disk laser. Engineering Review. 2011;31(2): 125–132.

[159] Caiazzo F, Alfieri V, Cardaropoli F, Sergi V. Butt autogenous laser welding of AA 2024 aluminium alloy thin sheets with a Yb:YAG disk laser. International Journal of Advanced Manufacturing Technology. 2013;67: 2157–2169.

[160] Jones M, Heurtier P, Desrayaud C, Montheillet F, Allehaux D, Driver J. Correlation between microstructure and microhardness in a friction stir welded 2024 aluminium alloy. Scripta Materialia. 2005;52(8): 693–697.

[161] Sutton M., Yang B, Reynolds a. ., Taylor R. Microstructural studies of friction stir welds in 2024-T3 aluminum. Materials Science and Engineering: A. 2002;323(1-2): 160–166.

[162] Wang SC, Lefebvre F, Yan JL, Sinclair I, Starink MJ. VPPA welds of Al-2024 alloys: Analysis and modelling of local microstructure and strength. Materials Science and Engineering: A. 2006;431(1-2): 123–136.

[163] Enz J, Khomenko V, Riekehr S, Ventzke V, Huber N, Kashaev N. Single-sided laser beam welding of a dissimilar AA2024–AA7050 T-joint. Materials & Design. 2015;76: 110–116.

[164] Wang J, Wang H-P, Wang X, Cui H, Lu F. Statistical analysis of process parameters to eliminate hot cracking of fiber laser welded aluminum alloy. Optics & Laser Technology. Elsevier, 2015;66: 15–21.

[165] Wang X, Lu F, Wang H-P, Cui H, Tang X, Wu Y. Mechanical constraint intensity effects on solidification cracking during laser welding of aluminum alloys. Journal of Materials Processing Technology. 2015;218: 62–70.

[166] Passini A, Oliveira AC De, Riva R, Travessa DN, Cardoso KR. Ultrasonic inspection of AA6013 laser welded joints. Materials Research. 2011;14(3): 417–422.

[167] Lima MS, Riva R, Oliveira AC De, Siqueira GR. Laser beam welding aerospace aluminum using fiber lasers. In: Dreischuh T, Atanasov PA, Sabotinov N V. (eds.) Proceedings of the 17th International Symposium on Gas Flow and Chemical Lasers and High Power Lasers. International Society for Optics and Photonics.; 2008. p. 713128.

269 [168] Katayama S, Nagayama H, Mizutani M, Kawahito Y. Fibre laser welding of aluminium alloy. Welding International. 2009;23(10): 744–752.

[169] Oliveira AC, Siqueira RHM, Riva R, Lima MSF. One-sided laser beam welding of autogenous T-joints for 6013-T4 aluminium alloy. Materials & Design. 2015;65: 726– 736.

[170] de Siqueira RHM, de Oliveira AC, Riva R, Abdalla AJ, Baptista CARP, de Lima MSF. Mechanical and microstructural characterization of laser-welded joints of 6013-T4 aluminum alloy. Journal of the Brazilian Society of Mechanical Sciences and Engineering. 2015;37(1): 133–140.

[171] Wang XJ, Lu FG, Wang HP, Cui HC, Tang XH, Wu YX. Experimental and numerical analysis of solidification cracking behaviour in fibre laser welding of 6013 aluminium alloy. Science and Technology of Welding and Joining. 2015;20(1): 58–67.

[172] Gao M, Chen C, Hu M, Guo L, Wang Z, Zeng X. Characteristics of plasma plume in fiber laser welding of aluminum alloy. Applied Surface Science. 2015;326: 181–186.

[173] Kawahito Y, Matsumoto N, Abe Y, Katayama S. Laser absorption of aluminium alloy in high brightness and high power fibre laser welding. Welding International. 2012;26(4): 275–281.

[174] Kawahito Y, Matsumoto N, Abe Y, Katayama S. Laser absorption characteristics in high-power fibre laser welding of stainless steel. Welding International. 2013;27(2): 129–135.

[175] Kawahito Y, Matsumoto N, Abe Y, Katayama S. Relationship of laser absorption to keyhole behavior in high power fiber laser welding of stainless steel and aluminum alloy. Journal of Materials Processing Technology. 2011;211(10): 1563–1568.

[176] Yu Y, Wang C, Hu X, Wang J, Yu S. Porosity in fiber laser formation of 5A06 aluminum alloy. Journal of Mechanical Science and Technology. 2010;24(5): 1077– 1082.

[177] Paleocrassas AG, Tu JF. Low-speed laser welding of aluminum alloy 7075-T6 using a 300-W, single-mode, ytterbium fiber laser. Welding Journal. 2007;86: s179–s186.

[178] Zhang ZH, Dong SY, Wang YJ, Xu BS, Fang JX, He P. Study on microstructures and mechanical properties of super narrow gap joints of thick and high strength aluminum alloy plates welded by fiber laser. The International Journal of Advanced Manufacturing Technology. 2015;

270 [179] Zhang Z, Dong S, Wang Y, Xu B, Fang J, He P. Microstructure characteristics of thick aluminum alloy plate joints welded by fiber laser. Materials & Design. 2015;84: 173– 177.

[180] Allen CM, Verhaeghe G, Hilton PA, Heason CP, Prangnell PB. Laser and Hybrid Laser-MIG Welding of 6.35 and 12.7mm Thick Aluminium Aerospace Alloy. Materials Science Forum. 2006;519-521: 1139–1144.

[181] Brown RT. Keyhole welding studies with a moderate-power, high-brightness fiber laser. Journal of Laser Applications. 2008;20(4): 201.

[182] Shiganov IN, Kholopov AA, Ioda EI. Special features of laser welding of aluminium alloys. Welding International. 2012;26(3): 231–235.

[183] Li Z, Gobbi SL, Norris I, Zolotovsky S, Richter KH. Laser Welding Techniques for Titanium Alloy Sheet. Journal of Materials Processing Technology. 1997;65: 203–208.

[184] Yunlian Q, Ju D, Quan H, Liying Z. Electron beam welding, laser beam welding and gas tungsten arc welding of titanium sheet. Materials Science and Engineering: A. 2000;280(1): 177–181.

[185] Wang SHH, Wei MDD, Tsay LWW. Tensile properties of LBW welds in Ti–6Al–4V alloy at evaluated temperatures below 450 °C. Materials Letters. 2003;57(12): 1815– 1823.

[186] Caiazzo F, Curcio F, Daurelio G, Memola Capece Minutolo F. Ti6Al4V sheets lap and butt joints carried out by CO2 laser: mechanical and morphological characterization. Journal of Materials Processing Technology. 2004;149(1-3): 546–552.

[187] Casalino G, Curcio F, Memola Capece Minutolo F. Investigation on Ti6Al4V laser welding using statistical and Taguchi approaches. Journal of Materials Processing Technology. 2005;167(2-3): 422–428.

[188] Chen L, Hu L, Gong S. A study on the porosity of CO2 laser welding of titanium alloy. China Welding. 2006;15(1): 1–5.

[189] Casavola C, Pappalettere C, Tattoli F. Experimental and numerical study of static and fatigue properties of titanium alloy welded joints. Mechanics of Materials. 2009;41(3): 231–243.

[190] Cao X, Jahazi M. Effect of welding speed on butt joint quality of Ti–6Al–4V alloy welded using a high-power Nd:YAG laser. Optics and Lasers in Engineering. 2009;47(11): 1231–1241.

271 [191] Kabir ASH, Cao X, Medraj M, Wanjara P, Cuddy J, Birur A. Effect of Welding Speed and Defocusing Distance on the Quality of Laser Welded Ti-6Al-4V. Proceedings of the Materials Science and Technology (MS&T) 2010 Conference. Houston, TX: ASM International; 2010. p. 2787–2797.

[192] Serroni G, Bitondo C, Squillace A. Mechanical characterization of Nd : YAG laser welding on Ti-6Al-4V. Innovative Production Machines and System. 2010;

[193] Squillace A, Prisco U, Ciliberto S, Astarita A. Effect of welding parameters on morphology and mechanical properties of Ti-6Al-4V laser beam welded butt joints. Journal of Materials Processing Technology. 2012;212(2): 427–436.

[194] Zhao S, Yu G, He X, Hu Y. Microstructural and mechanical characteristics of laser welding of Ti6Al4V and lead metal. Journal of Materials Processing Technology. 2012;212(7): 1520–1527.

[195] Scintilla LD, Palumbo G, Sorgente D, Tricarico L. Fiber laser cutting of Ti6Al4V sheets for subsequent welding operations: Effect of cutting parameters on butt joints mechanical properties and strain behaviour. Materials & Design. 2013;47: 300–308.

[196] Blackburn JE, Allen CM, Hilton P a., Li L, Hoque MI, Khan a. H. Modulated Nd : YAG laser welding of Ti–6Al–4V. Science and Technology of Welding and Joining. 2010;15(5): 433–439.

[197] Akman E, Demir a., Canel T, Sınmazçelik T. Laser welding of Ti6Al4V titanium alloys. Journal of Materials Processing Technology. 2009;209(8): 3705–3713.

[198] Fan Y, Shipway PH, Tansley GD, Xu J. The Effect of Heat Treatment on Mechanical Properties of Pulsed Nd:YAG Welded Thin Ti6Al4V. Advanced Materials Research. 2011;189-193: 3672–3677.

[199] Missori S, Murdolo F, Sili A. Laser beam welding of cast commercial pure titanium. Metallurgical Science and Technology. 2003;21(2): 26–30.

[200] XU P. Microstructure characterization of Ti–6Al–4V titanium laser weld and its deformation. Transactions of Nonferrous Metals Society of China. The Nonferrous Metals Society of China, 2012;22(9): 2118–2123.

[201] Jianxun Z, Xu S, Li Z. Investigation into plastic damage behavior of the CO2 laser deep penetration welded joint for Ti–6Al–4V alloy. Engineering Fracture Mechanics. 2012;83: 1–7.

[202] Balasubramanian TS, Balakrishnan M, Balasubramanian V, Muthu Manickam MA. Influence of welding processes on microstructure, tensile and impact properties of Ti-

272 6Al-4V alloy joints. Transactions of Nonferrous Metals Society of China (English Edition). The Nonferrous Metals Society of China, 2011;21(6): 1253–1262.

[203] Brandizzi M, Satriano AA, Sorgente D, Tricarico L. Laser–arc hybrid welding of Ti6Al4V titanium alloy: mechanical characterization of joints and gap tolerance. Welding International. 2011;27(2): 1–8.

[204] Mueller S, Stiles E, Dienemann R. Study of porosity formation during laser welding of Ti6Al4V. Proceedings of the 25th International Congress on Applications of Lasers and Electro-Optics. Scottsdale, AZ: Laser Institute of America; 2006. p. 304.

[205] Costa A, Miranda R, Quintino L, Yapp D. Analysis of Beam Material Interaction in Welding of Titanium with Fiber Lasers. Materials and Manufacturing Processes. 2007;22(7-8): 798–803.

[206] Fan Y, Chen Z, Zhang CH, Liu AM. A comparison of microstructure and mechanical properties of welded thin Ti6Al4V with three different types of laser. Materials Research Innovations. 2015;19(S4): S187–S192.

[207] Campanelli SL, Casalino G, Mortello M, Angelastro A, Ludovico AD. Microstructural Characteristics and Mechanical Properties of Ti6Al4V Alloy Fiber Laser Welds. Procedia CIRP. 2015;33: 428–433.

[208] Chen HC, Pinkerton AJ, Li L. Fibre laser welding of dissimilar alloys of Ti-6Al-4V and Inconel 718 for aerospace applications. The International Journal of Advanced Manufacturing Technology. 2010;52(9-12): 977–987.

[209] Li C, Muneharua K, Takao S, Kouji H. Fiber laser-GMA hybrid welding of commercially pure titanium. Materials & Design. 2009;30(1): 109–114.

[210] Liu D, Atkinson HV, Kapranos P, Jirattiticharoean W, Jones H. Microstructural evolution and tensile mechanical properties of thixoformed high performance aluminium alloys. Materials Science and Engineering: A. 2003;361(1-2): 213–224.

[211] Uspenskiy SA, Petrovskiy VN, Bykovskiy DP, Mironov VD, Prokopova NM, Tret’yakov E V. Spectral diagnostics of a vapor-plasma plume produced during welding titanium with a high-power ytterbium fiber laser. Journal of Physics: Conference Series. 2015;594(1): 012033.

[212] Clifford M. ASME engineer’s data book.. 1st ed. Fairfield: ASME Press; 2001. 211 p.

[213] Benyounis KY, Olabi AG, Hashmi MSJ. Effect of laser welding parameters on the heat input and weld-bead profile. Journal of Materials Processing Technology. 2005;164- 165: 978–985.

273 [214] El-Batahgy A-M. Effect of laser welding parameters on fusion zone shape and solidification structure of austenitic stainless steels. Materials Letters. 1997;32(2-3): 155–163.

[215] Gong S, Chen L, Yao W, Tang T. Research on laser welding of BT20 titanium alloy. In: Li L, Achara C (eds.) Proceedings of the 22nd International Congress on Application of Lasers & Electro-Optics. Jacksonville, Fl: Laser Institute of America; 2003. p. 1–6.

[216] Lampa C, Powell J, Ivarson A, Runnemalm H, Magnusson C. The influence of gap width on laser welding. Proceedings of the 14th International Congress on Applications of Lasers and Electro-optics. Orlando, Fl: Laser Institute of America; 1995. p. 504–512.

[217] Karlsson J, Markmann C, Minhaj Alam M, Kaplan AFH. Parameter influence on the laser weld geometry documented by the matrix flow chart. Physics Procedia. 2010;5: 183–192.

[218] Manonmani K, Murugan N, Buvanasekaran G. Effect of process parameters on the weld bead geometry of laser beam welded stainless steel sheets. International Journal for the Joining of Materials. 2005;17(4): 103–109.

[219] Zacharia T, David SA, Vitek JM, Debroy T. Weld pool development during GTA and laser beam welding of type 304 stainless steel, Part II—experimental correlation. Welding Journal. 1989;68(12): 510s – 519s.

[220] Coniglio N, Cross CE, Michael T, Lammers M. Defining a Critical Weld Dilution to Avoid Solidification Cracking in Aluminum. Welding Journal. 2008;87: 237–247.

[221] Coniglio N, Cross CE. Initiation and growth mechanisms for weld solidification cracking. International Materials Reviews. 2013;58(7): 375–397.

[222] Coniglio N, Cross CE. Mechanisms for solidification crack initiation and growth in aluminum welding. Metallurgical and Materials Transactions A. 2009;40(11): 2718– 2728.

[223] Dixon BF, Ritter JC. Solidification cracking in high strength steels welded by electron beam and laser beam processes. Proceedings of the 7th International Japanese Welding Society Symposium. Kobe: Japanese Welding Society; 2001.

[224] Vitek JM, Iskander S, Oblow EW. Neural network modeling of pulsed-laser weld pool shapes in aluminum alloy welds. Proceedings of 5th International Conference on

274 Trends in Welding Research. Pine Mountain, GA: ASM International; 1998. p. 442– 448.

[225] Lee HK, Han HS, Son KJ, Hong SB. Optimization of Nd:YAG laser welding parameters for sealing small titanium tube ends. Materials Science and Engineering: A. 2006;415(1-2): 149–155.

[226] Mueller S, Braat C, Mueller P, Cuddy J, Shankar K. Laser beam welding of titanium – a comparison of CO2 and fibre laser for potential aerospace applications. Proceedings of the 27th International Congress on Applications of Lasers and Electro- Optics. Temecula: Laser Institute of America; 2008. p. 1802.

[227] Hilton P, Blackburn J, Chong P. Welding of Ti-6Al-4V with fibre delivered laser beams. Proceedings of the 26th International Congress on Applications of Lasers and Electro- optics. Orlando, FL: Laser Institute of America; 2007. p. 887–895.

[228] Steen WM, Mazumder J. Basic Laser Optics. Laser Material Processing. 4th ed. London: Springer-Verlag London; 2010. p. 79–128.

[229] Nesterov A V, Niziev VG. Laser beams with axially symmetric polarization. Journal of Physics D: Applied Physics. 2000;33(15): 1817–1822.

[230] Cho J-H, Farson DF, Reiter MJ. Analysis of penetration depth fluctuations in single- mode fibre laser welds. Journal of Physics D: Applied Physics. 2009;42(11): 115501.

[231] BS EN 4678:2011 Aerospace series. Weldments and brazements for aerospace structures. Joints of metallic materials by laser beam welding. Quality of weldments.

[232] Cao X, Jahazi M, Immarigeon JP, Wallace W. A review of laser welding techniques for magnesium alloys. Journal of Materials Processing Technology. 2006;171(2): 188–204.

[233] ISO/TR17671-6:2005 Welding - Recommendations for welding of metallic materials - Part 6:Laser beam welding.

[234] AWS C7.2 Recommended Practices for Laser Beam Welding, Cutting and Drilling.

[235] BS EN 1011-6:2005 Welding. Recommendations for welding of metallic materials. Laser beam welding.

[236] ASTM E3 Methods for Preparation of Metallographic Specimens.

[237] BS EN 1321:1997 Destructive test on welds in metallic materials. Macroscopic and microscopic examination of welds.

275 [238] Peters M, Hemptenmacher J, Kumpfert J, Leyens C. Structure and Properties of Titanium and Titanium Alloys. In: Leyens C, Peters M (eds.) Titanium and Titanium Alloys. Fundamentals and Applications. 1st ed. Weinheim: WILEY-VCH Verlag GmbH & Co. KGaA; 2003. p. 1–36.

[239] BS EN ISO 15614-11:2002 Specification and qualification of welding procedures for metallic materials — Welding procedure test — Part 11 Electron and laser beam welding.

[240] AWS D17.1:2001 Specification for Fusion Welding for Aerospace Applications.

[241] BS EN ISO 6520-1:2007 Welding and allied processes — Classification of geometric imperfections in metallic materials — Part 1 Fusion welding.

[242] Yang JG, Sung SH, Chen CS, Tan AH. Study of microstructural and mechanical properties of weld heat affected zones of 2024-T3 aluminium using Gleeble simulation. Materials Science and Technology. 2011;27(1): 357–365.

[243] Wang L, Strangwood M, Balint D, Lin J, Dean TA. Formability and failure mechanisms of AA2024 under hot forming conditions. Materials Science and Engineering: A. 2011;528(6): 2648–2656.

[244] Ma T, den Ouden G. Softening behaviour of Al–Zn–Mg alloys due to welding. Materials Science and Engineering: A. 1999;266(1-2): 198–204.

[245] Watkins K., McMahon M., Steen W. Microstructure and corrosion properties of laser surface processed aluminium alloys: a review. Materials Science and Engineering: A. 1997;231(1-2): 55–61.

[246] Dudas JH, Collins RR. Preventing Weld Cracks in High-Strength Aluminum Alloys. Welding Journal. 1966;45: 241s – 249s.

[247] Kaplan A. Metallurgy of Welding and Hardening. In: Dowden J (ed.) The Theory of Laser Materials Processing: Heat and Mass Transfer in Modern Technology. 1st ed. Briston: Canopus Academic Publishing; 2009. p. 217–233.

[248] Huang C, Kou S. Liquation Cracking in Full-penetration Al-CU Welds. Welding Journal. 2004;83(2): 111S – 122S.

[249] Koseki T, Inoue H, Fukuda Y, Nogami A. Numerical simulation of equiaxed grain formation in weld solidification. Science and Technology of Advanced Materials. 2003;4(2): 183–195.

276 [250] Schempp P, Cross CE, Pittner A, Oder G, Neumann RS, Rooch H, et al. Solidification of GTA aluminum weld metal: Part 1 - Grain morphology dependent upon alloy composition and grain refiner content. Welding journal. 2014;93(2): 53s – 59s.

[251] Migliore L. Laser-material interactions. Manufacturing Engineering and Materials Processing. 1996;46: 65–82.

[252] Dawes C. Laser Welding: A practical guide.. 1st ed. Cambridge: Woodhead Publishing; 1992. 272 p.

[253] Naeem M, Jessett R. Welding aluminum tailored blanks with Nd:YAG lasers for automotive applications. Practical Welding Today. 1999;3(1): 24–28.

[254] Lanzafame J, Kattamis T. Solidification structure and tensile properties of 2014 aluminum alloy welds. Welding Journal. 1973;52(5): 226s – 232s.

[255] Pastor M, Zhao H, Martukanitz RP, DebRoy T. Porosity, Underfill and Magnesium loss during continuous wave Nd: YAG laser welding of thin plates of aluminum alloys 5182 and 5754. Welding Journal. 1999;78(6): 207s – 216s.

[256] Alfieri V, Cardaropoli F, Caiazzo F, Sergi V. Investigation on Porosity Content in 2024 Aluminum Alloy Welding by Yb:YAG Disk Laser. Advanced Materials Research. 2011;383-390: 6265–6269.

[257] Haboudou A, Peyre P, Vannes AB, Peix G. Reduction of porosity content generated during Nd:YAG laser welding of A356 and AA5083 aluminium alloys. Materials Science and Engineering: A. 2003;363(1-2): 40–52.

[258] Zhao H, DebRoy T. Macroporosity free aluminum alloy weldments through numerical simulation of keyhole mode laser welding. Journal of Applied Physics. 2003;93(12): 10089.

[259] Davis JR. Fabrication and Finishing of Aluminum Alloys. In: Davis JR (ed.) Aluminum and aluminum alloys. ASM International; 1993. p. 199–451.

[260] Zhou J, Tsai H-L. Porosity Formation and Prevention in Pulsed Laser Welding. Journal of Heat Transfer. 2007;129(8): 1014.

[261] Cross CE. On the Origin of Weld Solidification Cracking. In: Böllinghaus T, Herold H (eds.) Hot Cracking Phenomena in Welds. Berlin/Heidelberg: Springer-Verlag; 2005. p. 3–18.

[262] Klein T, Vicanek M, Kroos J, Decker I, Simon G. Oscillations of the keyhole in penetration laser beam welding. Journal of Physics D: Applied Physics. 1994;27(10): 2023–2030.

277 [263] Abbaschian L, Lima MSF De. Cracking susceptibility of aluminum alloys during laser welding. Materials Research. 2003;6(2): 273–278.

[264] Hu B, Richardson IM. Mechanism and possible solution for transverse solidification cracking in laser welding of high strength aluminium alloys. Materials Science and Engineering: A. 2006;429(1-2): 287–294.

[265] Salminen A. The Filler Wire - Laser Beam Interaction during Laser Welding with Low Alloyed Steel Filler Wire. Mechanika. 2010;84(4): 67–74.

[266] ARATA Y, MARUO H, MIYAMOTO I, NISHIO R. High Power CO2 Laser Welding of Thick Plate : Multipass Weding with Filler Wire(Welding Physics, Process & Instrument). Transactions of JWRI. 1986;15(2): 199–206.

[267] Takahashi K, Mehmetli B, Sato S. Influence of shielding gas and laser irradiation conditions on porosity formation in CO 2 laser welding of aluminium alloy. Welding International. 1998;12(5): 347–353.

[268] Pastor M, Zhao H, Debroy T. Pore formation during continuous wave Nd:YAG laser welding of aluminium for automotive applications. Welding International. p, 2001;15(4): 275–281.

[269] Ghaznavi MM, Ekrami A, Kokabi AH. Effect of solidification mechanism on microstructure and mechanical properties of joint in TLP bonded Al2024‐ T6 alloy. Science and Technology of Welding and Joining. 2011;16(2): 174–180.

[270] Hutchinson CR, Ringer SP. Precipitation processes in Al-Cu-Mg alloys microalloyed with Si. Metallurgical and Materials Transactions A. 2000;31(11): 2721–2733.

[271] Michaud EJ, Kerr HW, Weckman DC. Temporal pulse shaping and solidification cracking in laser welded Al-Cu alloys. Proceedings of the 4th International Conference on Trends in Welding Research. Gatlinburg, TN: ASM International; 1995. p. 153– 158.

[272] Schubert E, Klassen M, Zerner I, Walz C, Sepold G. Light-weight structures produced by laser beam joining for future applications in automobile and aerospace industry. Journal of Materials Processing Technology. 2001;115(1): 2–8.

[273] Akhter R, Davis M, Dowden J, Kapadia P, Ley M, Steen WM. A method for calculating the fused zone profile of laser keyhole welds. Journal of Physics D: Applied Physics. 1989;22(1): 23–28.

[274] Kim KR, Farson DF. CO2 laser–plume interaction in materials processing. Journal of Applied Physics. 2001;89(1): 681.

278 [275] Kawahito Y, Kinoshita K, Matsumoto N, Mizutani M, Katayama S. Effect of weakly ionised plasma on penetration of stainless steel weld produced with ultra high power density fibre laser. Science and Technology of Welding and Joining. 2008;13(8): 744– 748.

[276] Kawahito Y, Matsumoto N, Mizutani M, Katayama S. Characterisation of plasma induced during high power fibre laser welding of stainless steel. Science and Technology of Welding and Joining. 2008;13(8): 749–753.

[277] Greses J, Barlow CY, Steen WM, Hilton PA. Spectroscopic studies of plume/plasma in different gas environments. Proceedings of the 20th International Congress on Applications of Lasers & Electro-Optics. Jacksonville, Fl: Laser Institute of America; 2001. p. 808.

[278] Xiao R, Zuo T, Volz R. Experimental research on the influence of laser-induced plasma on the beam focusing during high-power CO2 laser materials processing. Chinese Journal of Lasers B. 1998;B7(6): 550–554.

[279] David SA, Vitek JM. Correlation between solidification parameters and weld microstructures. International Materials Reviews. Maney PublishingSuite 1C, Joseph’s Well, Hanover Walk, Leeds LS3 1AB, UK, 2013;

[280] David SA, Babu SS, Vitek JM. Welding: Solidification and microstructure. JOM. 2003;55(6): 14–20.

[281] Elmer JW, Palmer TA, Babu SS, Zhang W, DebRoy T. Phase transformation dynamics during welding of Ti-6Al-4V. Journal of Applied Physics. 2004;95: 8327– 8339.

[282] Ahmed T, Rack HJ. Phase transformations during cooling in α+β titanium alloys. Materials Science and Engineering: A. 1998;243: 206–211.

[283] Babu B. Physically Based Model for Plasticity and Creep of Ti-6Al-4V. [Online] Luleå University of Technology; 2008.

[284] Kabir ASH. Nd:YAG Laser Welding of Ti-6Al-4V Alloy. Concordia University; 2011.

[285] Murav’ev VI. Problems of Pore Formation in Welded Joints of Titanium Alloys. Metal Science and Heat Treatment. 2005;47(7-8): 282–288.

[286] Khaled T. An investigation of pore cracking in titanium welds. Journal of Materials Engineering and Performance. 1994;3(1): 21–36.

[287] Mitchell DR. Porosity in Titanium Welds. Welding Journal. 1982;61: 157s – 167s.

279 [288] Verhaeghe G, Hilton P, Barnes S. Achieving Low-Porosity Laser Welds in Aerospace Aluminium Alloy. 2003 Aerospace Manufacturing Technology Conference. Montreal: SAE International; 2003. p. 2003–01 – 2895.

[289] Weeter L. Technological advances in aluminum laser welding. Practical Welding Today. 1998;2(1): 56–57.

[290] Vitek JM, Babu SS. Multiscale characterisation of weldments. Science and Technology of Welding and Joining. 2011;16(1): 3–11.

[291] Tong W, Tao H, Zhang N, Jiang X, Marya MP, Hector LG, et al. Deformation and fracture of miniature tensile bars with resistance-spot-weld microstructures. Metallurgical and Materials Transactions A. 2005;36(10): 2651–2669.

[292] Scheider I, Brocks W, Cornec A. Procedure for the Determination of True Stress- Strain Curves From Tensile Tests With Rectangular Cross-Section Specimens. Journal of Engineering Materials and Technology. American Society of Mechanical Engineers, 2004;126(1): 70.

[293] Lockwood W, Tomaz B, Reynolds AP. Mechanical response of friction stir welded AA2024: experiment and modeling. Materials Science and Engineering A. 2002;323: 348–353.

[294] Reynolds AP, Duvall F. Digital image correlation for determination of weld and base metal constitutive behavior. Welding Research Supplement. 1999;78: 355–360.

[295] Pan B, Qian K, Xie H, Asundi A. Two-dimensional digital image correlation for in- plane displacement and strain measurement: a review. Measurement Science and Technology. 2009;20: 062001.

[296] Tung S-H, Shih M-H, Kuo J-C. Application of digital image correlation for anisotropic plastic deformation during tension testing. Optics and Lasers in Engineering. Elsevier, 2010;48(5): 636–641.

[297] Chevalier L, Calloch S, Hild F, Marco Y. Digital image correlation used to analyze the multiaxial behavior of rubber-like materials. European Journal of Mechanics - A/Solids. 2001;20(2): 169–187.

[298] Lecompte D, Sol H, Vantomme J, Habraken A. Analysis of speckle patterns for deformation measurements by digital image correlation. In: Slangen P, Cerruti C (eds.) Proceedings of SPIE Speckle06: Speckles, From Grains to Flowers. Nimes: International Society for Optics and Photonics; 2006. p. 63410E.

280 [299] Hung P-C, Voloshin AS. In-plane strain measurement by digital image correlation. Journal of the Brazilian Society of Mechanical Sciences and Engineering. The Brazilian Society of Mechanical Sciences and Engineering, 2003;25(3): 215–221.

[300] ASTM E384 Standard Test Method for Micro-indentation Hardness of Materials.

[301] BS ISO 22826:2005 Destructive tests on welds in metallic materials. Hardness testing of narrow joints welded by laser and electron beam (Vickers and Knoop hardness tests).

[302] ASTM E8 Standard Methods for Tension Testing of Metallic Materials.

[303] BS EN ISO 6892-1:2009 Metallic materials. Tensile testing. Method of test at ambient temperature.

[304] AWS B4.0:2007 Standard Methods for Mechanical Testing of Welds.

[305] Sun Z, Lyons JS, McNeill SR. Measuring Microscopic Deformations with Digital Image Correlation. Optics and Lasers in Engineering. 1997;27(4): 409–428.

[306] Kang J, Jain M, Wilkinson DS, Embury JD. Microscopic Strain Mapping Using Scanning Electron Microscopy Topography Image Correlation at Large Strain. The Journal of Strain Analysis for Engineering Design. 2005;40(6): 559–570.

[307] Ahn B, Nutt SR. Strain Mapping of Al–Mg Alloy with Multi-scale Grain Structure using Digital Image Correlation Method. Experimental Mechanics. 2009;50(1): 117–123.

[308] Preston RV, Shercliff HR, Withers PJ, Smith S. Physically-based constitutive modelling of residual stress development in welding of aluminium alloy 2024. Acta Materialia. 2004;52(17): 4973–4983.

[309] Katoh M. Factors affecting mechanical properties of laser welded aluminium alloys. Welding International. 1996;10(10): 771–777.

[310] ASME B31.1:1935 Code for Pressure Piping.

[311] Bouafia F, Serier B, Serier N, Hayat SS. Effect of Density and Pointed Corner Degree of Pore on Local Stress in Welded Structures: Defect in Marine Structures. International Scholarly Research Notices Mechanical Engineering. 2014;2014(834659): 1–7.

[312] Ludovico AD, Daurelio G, De Filippis LAC, Scialpi A, Squeo F. Laser welding of the AA 2024-T3 aluminum alloy by using two different laser sources (Nd:YAG or CO2). In: Kodymova J (ed.) Proceedings of the XV International Symposium on Gas Flow, Chemical Lasers, and High-Power Lasers. 2005. p. 887–894.

281 [313] Chang WS, Rajesh SR, Chun CK, Kim HJ. Microstructure and mechanical Properties of Hybrid Laser-friction Stir Welding Between AA6061-T6 Al Alloy and AZ31 Mg Alloy. Journal of Materials Science and Technology. 2011;27(3): 199–204.

[314] Chao YJ, Qi X, Tang W. Heat Transfer in Friction Stir Welding – Experimental and numerical studies. Journal of Manufacturing Science and engineering. 2003;125: 138–145.

[315] Flansbjer M, Sjögren T. Using Digital Image Correlation Techniques and Finite Element Models for Strain-Field Analysis of a Welded Aluminium Structure. Applied Mechanics and Materials. 2011;70: 123–128.

[316] Shore RJ, McCauley RB. Effects of Porosity on High Strength Aluminum 7039. Welding Journal Supplement. 1970;7: s311–s321.

[317] Molak RM, Paradowski K, Brynk T, Ciupinski L, Pakiela Z, Kurzydlowski KJ. Measurement of mechanical properties in a 316L stainless steel welded joint. International Journal of Pressure Vessels and Piping. 2009;86(1): 43–47.

[318] Genevois C, Deschamps A, Vacher P. Comparative study on local and global mechanical properties of 2024 T351, 2024 T6 and 5251 O friction stir welds. Materials Science and Engineering A. 2006;415: 162–170.

[319] Padmanabham G, Shanmugarajan B. Experimental Investigation of Bead-on-Bead CO2 Laser Welding of Al Alloy 6061. In: David SA, DebRoy T, DuPont JN, Koseki T, Smartt HB (eds.) Trends in Welding Research 2008: Proceedings of the 8th International Conference. Pine Mountain, GA: ASM International; 2009. p. 598–603.

[320] Kabir ASH, Cao X, Gholipour J, Wanjara P, Cuddy J, Birur A, et al. Effect of Postweld Heat Treatment on Microstructure, Hardness, and Tensile Properties of Laser-Welded Ti-6Al-4V. Metallurgical and Materials Transactions A. 2012;43(11): 4171–4184.

[321] Cao X, Kabir ASH, Wanjara P, Gholipour J, Birur A, Cuddy J, et al. Global and Local Mechanical Properties of Autogenously Laser Welded Ti-6Al-4V. Metallurgical and Materials Transactions A. 2014;45(3): 1258–1272.

[322] Abu SHK, Cao XJ, Gholipour Baradari J, Wanjara P, Cuddy J, Birur A, et al. Determination of Global and Local Tensile Behaviours of Laser Welded Ti-6Al-4V Alloy. Advanced Materials Research. 2011;409: 859–864.

[323] Oh J, Kim NJ, Lee S, Lee EW. Correlation of fatigue properties and microstructure in investment cast Ti-6Al-4V welds. Materials Science and Engineering: A. 2003;340(1- 2): 232–242.

282 [324] Moiseev VN, Polyak E V., Sokolova a. Y. Martensite strengthening of titanium alloys. Metal Science and Heat Treatment. 1975;17(8): 687–691.

[325] Fan Y. Mechanical Properties of Laser Beam Welded Ti6Al4V. [Online] University of Nottingham; 2010.

[326] Leitão C, Galvão I, Leal RM, Rodrigues DM. Determination of local constitutive properties of aluminium friction stir welds using digital image correlation. Materials and Design. 2012;33: 69–74.

[327] Kleemola HJ, Nieminen M a. On the strain-hardening parameters of metals. Metallurgical Transactions. 1974;5(8): 1863–1866.

[328] Gil FJ, Ginebra MP, Manero JM, Planell JA. Formation of α-Widmanstätten structure: effects of grain size and cooling rate on the Widmanstätten morphologies and on the mechanical properties in Ti6Al4V alloy. Journal of Alloys and Compounds. 2001;329(1-2): 142–152.

[329] Fan Y, Cheng P, Yao YL, Yang Z, Egland K. Effect of phase transformations on laser forming of Ti-6Al-4V alloy. Journal of Applied Physics. 2005;98: 1–10.

[330] Cho JY. Finite element modeling of martensitic phase transformation. Texas Tech University; 2009.

[331] Gil Mur FX, Rodríguez D, Planell JA. Influence of tempering temperature and time on the α′-Ti-6Al-4V martensite. Journal of Alloys and Compounds. 1996;234: 287–289.

[332] Masubuchi K. Fundamental Information on Residual Stresses. Analysis of Welded Structures: Residual Stresses, Distortion, and their Consequences. Oxford: Pergamon Press Ltd.; 1980. p. 94–111.

[333] Goldak JA, Akhlaghi M. Introduction. Computational Welding Mechanics. 1st ed. New York: Springer Science Business Media Inc.; 2005. p. 9–15.

[334] Flores-Johnson EA, Muránsky O, Hamelin CJ, Bendeich PJ, Edwards L. Numerical analysis of the effect of weld-induced residual stress and plastic damage on the ballistic performance of welded steel plate. Computational Materials Science. 2012;58: 131–139.

[335] Kou S. Residual Stresses, Distortion, and Fatigue. Welding Metallurgy. 2nd ed. Hoboken: John Wiley & Sons, Inc.; 2003. p. 123–127.

[336] Park MK, Sindhu RA, Lee SJ, Zai BA, Mehboob H. A Residual Stress Evaluation in Laser Welded Lap Joint with Hole Drilling Method. International Journal of Precision Engineering and Manufacturing. 2009;10(5): 89–95.

283 [337] James MN, Webster PJ, Hughes DJ, Chen Z, Ratel N, Ting S-P, et al. Correlating weld process conditions, residual strain and stress, microstructure and mechanical properties for high strength steel—the role of neutron diffraction strain scanning. Materials Science and Engineering: A. 2006;427(1-2): 16–26.

[338] DebRoy T, David SA. Physical processes in fusion welding.pdf. Reviews of Modern Physics. 1995;67(1): 85.

[339] Kannatey-Asibu E. Heat and fluid flow. Principles of Laser Materials Processing. New Jersey: John Wiley & Sons, Inc.; 2009. p. 231–232.

[340] Masubuchi K. Distortion in Weldments. Analysis of Welded Structures: Residual Stresses, Distortion, and their Consequences. Oxford: Pergamon Press Ltd.; 1980. p. 235–327.

[341] Kannatey-Asibu E. Residual Stresses and Distortion. Principles of Laser Materials Processing. New Jersey: John Wiley & Sons, Inc.; 2009. p. 361–405.

[342] Withers PJ, Bhadeshia HKDH. Residual stress. Part 1 – Measurement techniques. Materials Science and Technology. 2001;17(4): 355–365.

[343] Allen AJ, Hutchings MT, Windsor CG, Andreani C. Neutron diffraction methods for the study of residual stress fields. Advances in Physics. 1985;34(4): 445–473.

[344] Staron P, Vaidya WV, Koçak M, Homeyer J, Hackius J. Residual Stresses in Laser Beam Welded Butt Joints of the Airframe Aluminium Alloy AA6056. Materials Science Forum. 2006;524-525: 413–418.

[345] Ivetic G, Lanciotti A, Polese C. Electric strain gauge measurement of residual stress in welded panels. The Journal of Strain Analysis for Engineering Design. 2009;44(1): 117–126.

[346] Masubuchi K. Measurement of residual stresses in weldments. Analysis of Welded Structures: Residual Stresses, Distortion, and their Consequences. Oxford: Pergamon Press Ltd.; 1980. p. 112–147.

[347] Lorentzen T, Hutchings MT, Withers PJ, Holden TM. Introduction. Introduction to the characterization of residual stress by neutron diffraction. Florida: CRC Press; 2005. p. 1–24.

[348] Lorentzen T, Hutchings MT, Withers PJ, Holden TM. Interpretation and Analysis of Lattice Strain Data. Introduction to the characterization of residual stress by neutron diffraction. Florida: CRC Press; 2005. p. 213–273.

284 [349] Hutchings MT, Withers.P.J, Holden TM, Lorentzen T. Introduction to characterization of residual stress by neutron diffraction. Materials Today. 2005;8(5): 57.

[350] Eigenmann B, Macherauch E. Röntgenographische Untersuchung von Spannungszuständen in Werkstoffen Teil 3. Mat.-wiss. u. Werkstofftech. 1996;27: 426–437.

[351] Lorentzen T, Hutchings MT, Withers PJ, Holden TM. Practical Aspects of Strain Measurement Using Neutron Diffraction. Introduction to the characterization of residual stress by neutron diffraction. Florida: CRC Press; 2005. p. 155–212.

[352] Hauk V. Neutron diffraction methods. Structural and Residual Stress Analysis by Nondestructive Methods: Evaluation – Application – Assessment. Amsterdan: Elsevier Science B.V.; 1997. p. 495–519.

[353] Belassel M, Pineault J, Brauss ME. Review of Residual Stress Determination and Exploitation Techniques Using X-Ray Diffraction Method. Materials Science Forum. 2006;524-525: 229–234.

[354] Pirling T, Bruno G, Withers PJ. SALSA: Advances in Residual Stress Measurement at ILL. Materials Science Forum. 2006;524-525: 217–222.

[355] Lorentzen T, Hutchings MT, Withers PJ, Holden TM. Diffraction Techniques and Instrument Design. Introduction to the characterization of residual stress by neutron diffraction. Florida: CRC Press; 2005. p. 67–155.

[356] Karadge M, Grant B, Bruno G, Santisteban JR, Withers PJ, Preuss M. A Comparison of Strain Measurements on an Inertia Friction Weld Using the ENGIN-X and SALSA Neutron Strain Mapping Instruments. Materials Science Forum. 2006;524-525: 393– 398.

[357] Daymond MR, Bourke MAM, Von Dreele RB, Clausen B, Lorentzen T. Use of Rietveld refinement for elastic macrostrain determination and for evaluation of plastic strain history from diffraction spectra. Journal of Applied Physics. 1997;82(4): 1554.

[358] Johnson MW, Daymond MR. An optimum design for a time-of-flight neutron diffractometer for measuring engineering stresses. Journal of Applied Crystallography. 2002;35(1): 49–57.

[359] Preuss M, Withers PJ, Pang JWL, Baxter GJ. Inertia welding nickel-based superalloy: Part II. Residual stress characterization. Metallurgical and Materials Transactions A. 2002;33(10): 3227–3234.

285 [360] Leggatt RH. Residual stresses in welded structures. International Journal of Pressure Vessels and Piping. 2008;85: 144–151.

[361] ISO/TTA 3:2001 Polycrystalline materials - Determination of residual stresses by neutron diffraction.

[362] Stanford N, Bate PS. The Martensitic Transformation Texture in Ti-6Al-4V. Materials Science Forum. Trans Tech Publications, 2005;495-497: 669–674.

[363] Lorentzen T, Hutchings MT, Withers PJ, Holden TM. Fundamentals of neutron diffraction. Introduction to the characterization of residual stress by neutron diffraction. Florida: CRC Press; 2005. p. 25–66.

[364] Ganguly S, Edwards L, Fitzpatrick ME. Problems in using a comb sample as a stress- free reference for the determination of welding residual stress by diffraction. Materials Science and Engineering A. 2011;528: 1226–1232.

[365] Bayraktar FS, Staron P, Koçak M, Schreyer A. Analysis of Residual Stress in Laser Welded Aerospace Aluminium T-Joints by Neutron Diffraction and Finite Element Modelling. Materials Science Forum. 2008;571-572: 355–360.

[366] Radaj D. Modelling of welding residual stresses and distortion. Welding residual stresses and distortion: Calculation and measurement. 2nd ed. Düsseldorf: DVS- Verlag GmbH; 2003. p. 100–272.

[367] Inoue T. Thermal-metallurgical-mechanuical interactions during welding. In: Zhili F (ed.) Processes and mechanisms of welding residual stress and distortion. 1st ed. Cambridge: Woodhead Publishing Limited; 2005. p. 99–125.

[368] Ueda Y, Kim YC. Measuring Theory of Three-dimensional Residual Stresses Using a Thinly Sliced Plate Perpendicular to Welded Line. Transactions of JWRI. 1985;14(2): 151–157.

[369] Hill MR, Nelson D V. The localized eigenstrain method for determination of triaxial residual stress in welds. ASME Publications PVP. 1998;373: 397–404.

[370] Dong Y, Hong JK, Tsai CL, Dong P. Finite Element Modeling of Residual Stresses in Austenitic Stainless Steel Pipe Girth Welds. Welding Journal. 1997;76(10): 442s – 449s.

[371] Lindgren L-E. Finite element modeling and simulation of welding part 1: Increased complexity. Journal of Thermal Stresses. 2001;24(2): 141–192.

[372] Lindgren L-E. Finite element modeling and simulation of welding. Part 2: Improved material modeling. Journal of Thermal Stresses. 2001;24(3): 195–231.

286 [373] Lindgren L-E. Finite element modeling and simulation of welding. Part 3: efficiency and integration. Journal of Thermal Stresses. 2001;24(4): 305–334.

[374] Yaghi A, Becker A. State of the Review – Welding Simulations using Finite Element Methods Report No. FENET-UNOTT-DLE-08. University of Nottingham; 2004 p. 1– 27.

[375] He X. Finite Element Analysis of Laser Welding: A State of Art Review. Materials and Manufacturing Processes. 2012;(January 2013): 120813105610000.

[376] Zhu XKK, Chao YJJ. Effects of temperature-dependent material properties on welding simulation. Computers & Structures. 2002;80(11): 967–976.

[377] Deng D, Kiyoshima S. Numerical simulation of residual stresses induced by laser beam welding in a SUS316 stainless steel pipe with considering initial residual stress influences. Nuclear Engineering and Design. 2010;240(4): 688–696.

[378] Apostol G, Solomon G, During I. Input parameters influence on the residual stress and distortions at laser welding using finite element analysis. UPB Scientific Bulletin, Series D: Mechanical Engineering. 2012;74(2): 153–164.

[379] Friedman E. Thermomechanical Analysis of the Welding Process Using the Finite Element Method. Journal of Pressure Vessel Technology. 1975;97(3): 206.

[380] Ferro P, Porzner H, Tiziani a, Bonollo F. The influence of phase transformations on residual stresses induced by the welding process—3D and 2D numerical models. Modelling and Simulation in Materials Science and Engineering. 2006;14: 117–136.

[381] Gery D, Long H, Maropoulos P. Effects of welding speed, energy input and heat source distribution on temperature variations in butt joint welding. Journal of Materials Processing Technology. 2005;167(2-3): 393–401.

[382] Michaleris P, Feng Z, Campbell G. Evaluation of 2D and 3D FEA models for predicting residual stress and distortion. ASME Publications PVP. American Society of Mechanical Engineers, 1997;347: 91–102.

[383] Oddy AS, Goldak JA, McDill JMJ. Transformation plasticity and residual stresses in single-pass repair welds. Journal of Pressure Vessel Technology. 1992;114(1): 33.

[384] Dong P. Residual stresses and distortions in welded structures: a perspective for engineering applications. Science and Technology of Welding and Joining. 2005;10(4): 389–398.

287 [385] Pietzsch R, Brzoza M, Kaymak Y, Specht E, Bertram A. Minimizing the Distortion of Steel Profiles by Controlled Cooling. Steel Research International. 2005;76(5): 399– 407.

[386] Saad G, Fayek SA, Fawzy A, Soliman HN, Mohammed G. Deformation characteristics of Al-4043 alloy. Materials Science and Engineering: A. 2010;527(4-5): 904–910.

[387] Boivineau M, Cagran C, Doytier D, Eyraud V, Nadal MH, Wilthan B, et al. Thermophysical properties of solid and liquid Ti-6Al-4V (TA6V) alloy. International Journal of Thermophysics. 2006;27(2): 507–529.

[388] Mills KC. Recommended values of thermophysical properties for selected commercial alloys. Cambridge: Woodhead Publishing; 2002.

[389] Rai R, Elmer JW, Palmer T a, DebRoy T. Heat transfer and fluid flow during keyhole mode laser welding of tantalum, Ti–6Al–4V, 304L stainless steel and vanadium. Journal of Physics D: Applied Physics. 2007;40(18): 5753–5766.

[390] BECK T, RAU K. Temperature measurement and control methods in TMF testing – a comparison and evaluation. International Journal of Fatigue. 2008;30(2): 226–233.

[391] Davies CM, Ahn J, Tsunori M, Dye D, Nikbin KM. The Influence of Pre-existing Deformation on GMA Welding Distortion in Thin Steel Plates. Journal of Materials Engineering and Performance. 2015;24(1): 261–273.

[392] Rosenthal D. The theory of moving sources of heat and its application to metal treatments. Transaction of the American Society of Mechanical Engineers. 1946;68: 849–865.

[393] Rykalin RR. Energy sources for welding. Welding in the World. 1974;12(9-10): 227– 247.

[394] Carslaw HS, Jaeger JC. Conduction of heat in solids.. 2nd ed. Oxford: Clarendon Press; 1959.

[395] Myers PS, Uyehara OA, Borman GL. Fundamentals of heat flow in welding. Welding Research Council Bulletin. 1967;123: 1.

[396] Rybicki EF, Schmueser DW, Stonesifer RW, Groom JJ, Mishler HW. A Finite Element Model for Residual Stresses and Deflections in Girth Butt Welded Pipes. Journal of Pressure Vessel Technology. 1978;100: 256–262.

288 [397] Pavelic V, Tanbakuchi R, Uyehara OA, Myers P. Experimental and computed temperature histories in gas tungsten arc welding of thin plates. Welding Journal Research Supplement. 1969;48: 295s – 305s.

[398] Goldak J, Chakravarti A, Bibby M. A New Finite Element Model for Welding Heat Sources. Metallurgical Transaction B. 1984;15B(June): 299–305.

[399] Fuerschbach PW. Measurement and prediction of energy transfer efficiency in laser beam welding. Welding Journal. 1996;75(January): 33–39.

[400] Bradáč J. Calibration of heat source model in numerical simulations of fusion welding. Machines, Technologies, Materials. 2013;(11): 9–12.

[401] Khan P, Debroy T. Alloying element vaporization and weld pool temperature during laser welding of AlSl 202 stainless steel. Metallurgical and Materials Transactions B. 1984;15(4): 641–644.

[402] Wu CS, Chen J, Zhang YM. Numerical analysis of both front- and back-side deformation of fully-penetrated GTAW weld pool surfaces. Computational Materials Science. 2007;39(3): 635–642.

[403] Siddique M, Abid M, Junejo HF, Mufti RA. 3-D Finite Element Simulation of Welding Residual Stresses in Pipe-Flange Joints: Effect of Welding Parameters. Materials Science Forum. 2005;490-491: 79–84.

[404] Yang J, Sun S, Brandt M, Yan W. Experimental investigation and 3D finite element prediction of the heat affected zone during laser assisted machining of Ti6Al4V alloy. Journal of Materials Processing Technology. 2010;210(15): 2215–2222.

[405] Argyris JH, Szimmat J, Willam KJ. Computational aspects of welding stress analysis. Computer Methods in Applied Mechanics and Engineering. 1982;33(1-3): 635–665.

[406] Mullins J, Gunnars J. Welding Simulation: Relationship Between Welding Geometry and Determination of Hardening Model. Proceedings of the ASME 2012 Pressure Vessels and Piping Conference. Toronto: ASME; 2012. p. 1219.

[407] Ogawa K, Chidwick LO, Kingston EJ, Dennis R, Bray D, Yanagida N. The Measurement and Modelling of Residual Stresses in a Stainless Steel Pipe Girth Weld. In: Nitzel ME (ed.) Proceedings of the ASME 2008 Pressure Vessels and Piping Conference. Chicago: ASME; 2008. p. 527–534.

[408] Dong P. Residual Stress Analyses of a Multi-Pass Girth Weld: 3-D Special Shell Versus Axisymmetric Models. Journal of Pressure Vessel Technology. 2001;123(2): 207.

289 [409] Dye D, Hunziker O, Roberts SM, Reed RC. Modeling of the mechanical effects induced by the tungsten inert-gas welding of the IN718 superalloy. Metallurgical and Materials Transactions A. 2001;32(7): 1713–1725.

[410] Owen RA, Preston R V, Withers PJ, Shercliff HR, Webster PJ. Neutron and synchrotron measurements of residual strain in TIG welded aluminium alloy 2024. Materials Science and Engineering A. 2003;346: 159–167.

[411] Lindgren LE, Runnemalm H, Näsström MO. Simulation of multipass welding of a thick plate. International journal for numerical methods in engineering. 1999;44(9): 1301– 1316.

[412] Schenk T, Richardson IM, Kraska M, Ohnimus S. A study on the influence of clamping on welding distortion. Computational Materials Science. 2009;45(4): 999– 1005.

[413] Malik AM, Qureshi EM, Dar NU, Khan I. Residual Stress Fields Under Different Clamping Conditions in Circumferentially Welded Thin-Walled Cylinders. Proceedings of the 16th International Conference on Nuclear Engineering. Florida: ASME; 2008. p. 607–615.

[414] Feng Z. Mitigating welding residual stress and distortion. In: Feng Z (ed.) Processes and mechanisms of welding residual stress and distortion. Cambridge: Woodhead Publishing Limited; 2005. p. 264–294.

[415] Dean SW, Croucher T. Minimizing Machining Distortion in Aluminum Alloys through Successful Application of Uphill Quenching—A Process Overview. Journal of ASTM International. 2009;6(7): 101770.

[416] Leblond J. Metallurgical and mechanical consequences of phase transforma-tions in numerical simulations of welding processes. In: Karlsson L (ed.) Modeling in Welding, Hot Powder Forming, and Casting. ASM International; 1997. p. 61–89.

[417] Yaghi AH, Hyde TH, Becker AA, Sun W. Finite element simulation of welding and residual stresses in a P91 steel pipe incorporating solid-state phase transformation and post-weld heat treatment. The Journal of Strain Analysis for Engineering Design. 2008;43: 275–293.

[418] Yaghi AH, Hyde TH, Becker AA, Sun W. Numerical simulation of P91 pipe welding including the effects of solid-state phase transformation on residual stresses. Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications. 2007;221: 213–224.

290 [419] Yaghi AH, Hyde TH, Becker AA, Williams JA, Sun W. Residual stress simulation in welded sections of P91 pipes. Journal of Materials Processing Technology. 2005;167: 480–487.

[420] Lee CH, Chang KH. Prediction of residual stresses in high strength carbon steel pipe weld considering solid-state phase transformation effects. Computers and Structures. Elsevier Ltd, 2011;89(1-2): 256–265.

[421] Deng D, Murakawa H. Prediction of welding residual stress in multi-pass butt-welded modified 9Cr-1Mo steel pipe considering phase transformation effects. Computational Materials Science. 2006;37: 209–219.

[422] Taljat B, Radhakrishnan B, Zacharia T. Numerical analysis of GTA welding process with emphasis on post-solidification phase transformation effects on residual stresses. Materials Science and Engineering: A. 1998;246: 45–54.

[423] Szkliniarz W, Smotka G. Analysis of volume effects of phase transformation in titanium alloys. Journal of Materials Processing Technology. 1995;53(3): 413–422.

[424] JMatPro version 8.0. [Online]

[425] Elmer JW, Palmer T a., Babu SS, Specht ED. In situ observations of lattice expansion and transformation rates of alpha and beta phases in Ti-6Al-4V. Materials Science and Engineering A. 2005;391: 104–113.

[426] Shah AK, Kulkarni SD, Gopinathan V, Krishnan R. Weld Heat-Affected Zone in Ti-6Al- 4V Alloy 1. Computer-Simulation of the Effect of Weld Variables on the Thermal Cycles in the HAZ. Welding Journal. 1995;74: S297–S304.

[427] Shah AK, Kulkarni SD, Gopinathan V, Krishnan R. Weld Heat-Affected Zone in Ti-6Al- 4V Alloy 2. Modeling and Experimental Simulation of Growth and Phase- Transformations. Welding Journal. 1995;74: S325–S338.

[428] Malinov S, Markovsky P, Sha W, Guo Z. Resistivity study and computer modelling of the isothermal transformation kinetics of Ti-6Al-4V and Ti-6Al-2Sn-4Zr-2Mo-0.08Si alloys. Journal of Alloys and Compounds. 2001;314: 181–192.

[429] Malinov S, Guo Z, Sha W, Wilson A. Differential scanning calorimetry study and computer modeling of β ⇒ α phase transformation in a Ti-6Al-4V alloy. Metallurgical and Materials Transactions A. 2001;32: 879–887.

[430] Avrami M. Kinetics of phase change, I. General theory. Journal of Chemical Physics. 1939;7: 1103–1112.

291 [431] Avrami M. Kinetics of phase change, II. Transformation-time relations for random distribution of nuclei. Journal of Chemical Physics. 1940;8: 212–224.

[432] Avrami M. Kinetics of phase change, III. Granulation, phase change and microstructure. Journal of Chemical Physics. 1941;9: 177–184.

[433] Kelly SM, Kampe SL. Microstructural evolution in Laser-Deposited multilayer Ti6Al-4V build: Part I. Microstructural characterization. Metallurgical and Materials Transactions A. 2004;35(June): 1861–1867.

[434] Kelly SM, Kampe SL. Microstructural evolution in laser-deposited multilayer Ti-6Al-4V builds: Part II. Thermal modeling. Metallurgical and Materials Transactions A. 2004;35(June): 1869–1879.

[435] Crespo A, Deus A, Vilar R. Modeling of Phase Transformations and Internal Stresses in Laser Powder Deposition. In: Vilar R, Conde O, Fajardo M, Silva LO, Pires M, Utkin A (eds.) Proceedings of the 17th International Symposium on Gas Flow, Chemical Lasers and High-Power Lasers. Lisboa: International Society for Optical Engineering; 2008. p. 713120–713120 – 10.

[436] Murgau CC, Pederson R, Lindgren LE. A model for Ti–6Al–4V microstructure evolution for arbitrary temperature changes. Modelling and Simulation in Materials Science and Engineering. 2012;20: 055006.

[437] Longuet A, Robert Y, Aeby-Gautier E, Appolaire B, Mariage JF, Colin C, et al. A multiphase mechanical model for Ti–6Al–4V: Application to the modeling of laser assisted processing. Computational Materials Science. 2009;46(3): 761–766.

[438] Ducato a., Fratini L, Micari F. Prediction of phase transformation of Ti-6Al-4V titanium alloy during hot-forging processes using a numerical model. Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications. 2013;228: 154–159.

[439] Castro R, Seraphin L. Contribution à l’étude métallographique et structurale de l’alliage de titane TA6V. mémoires scientifiques revue de métallurgie. 1966;63: 1025– 1058.

[440] Filip R, Kubiak K, Ziaja W, Sieniawski J. The effect of microstructure on the mechanical properties of two-phase titanium alloys. Journal of Materials Processing Technology. 2003;133(1-2): 84–89.

[441] Cardona A, Baufeld B, Biest O Van Der, Fachinotti VD, Cardona A, Baufeld B, et al. Evolution of microstructure during shaped metal deposition. In: Dvorkin E, Goldschmit

292 M, Storti M (eds.) Mecánica Computacional Vol XXIX. Buenos Aires: Asociacion Argentina de Mecanica Computacional; 2010. p. 4927–4934.

[442] Katzarov I, Malinov S, Sha W. Finite element modeling of the morphology of β to α phase transformation in Ti-6Al-4V alloy. Metallurgical and Materials Transactions A. 2002;33: 1027–1040.

[443] Kelly SM. Thermal and Microstructure Modeling of Metal Deposition Processes with Application to Ti-6Al-4V. [Online] Virginia Polytechnic Institute and State University; 2004.

[444] Christian JW. Formal Theory of transformation kinetics. Theory of phase transformations in metals and alloys - Part I. 3rd ed. Oxford: Pergamon Press Ltd.; 2002. p. 546–552.

[445] Rohde J, Jeppsson A. Literature review of heat treatment simulations with respect to phase transformation, residual stresses and distortion. Scandinavian Journal of Metallurgy. 2000;29: 47–62.

[446] Provost W. Effects of a stress relief heat treatment on the toughness of pressure vessel quality steels—Influence of the plate thickness. International Journal of Pressure Vessels and Piping. 1982;10(2): 93–118.

[447] Dong P, Song S, Zhang J. Analysis of residual stress relief mechanisms in post-weld heat treatment. International Journal of Pressure Vessels and Piping. 2014;122: 6–14.

[448] Li JY, Lu YM, Liu ZP, Zhang MM. Residual Stresses and Distortions in Aluminum Alloy Sheet after Welding under Pre-Tension. Materials Science Forum. 2009;618- 619: 45–48.

[449] Guan Q. A survey of development in welding stress and distortion controlling in aerospace manufacturing engineering in China. Welding in the World. 1999;43(1): 64–74.

[450] Richards DG, Prangnell PB, Williams SW, Withers PJ. Global mechanical tensioning for the management of residual stresses in welds. Materials Science and Engineering: A. 2008;489(1-2): 351–362.

[451] Altenkirch J, Steuwer A, Peel M, Richards DG, Withers PJ. The effect of tensioning and sectioning on residual stresses in aluminium AA7749 friction stir welds. Materials Science and Engineering: A. 2008;488(1-2): 16–24.

[452] Yang Y-S, Lee S-H. A study on the mechanical stress relieving in a butt-welded-pipe. International Journal of Pressure Vessels and Piping. 1997;73(3): 175–182.

293 [453] Williams SW, Price DA, Wescott A, Harrison CJC, Staron P, Kocak R. Welding and Brazing of Aerospace Structures—Modern Applications and Materials for New and In- Service Parts. Proceedings of the Distortion Control in Welding by Stress Engineering. Berlin: DVS Berichte; 2004. p. 95–101.

[454] Qi XL, Zhang Y. Analysis on the Critical Buckling Stress in the TC4 Thin Plate Weldment. Materials Science Forum. 2011;704-705: 1387–1392.

[455] Xu J, Chen L, Ni C. A study on the mechanical stress relieving and safety assessment without post-weld heat treatment. Materials Science and Engineering: A. 2007;443(1- 2): 107–113.

[456] Richards DG, Prangnell PB, Withers PJ, Williams SW, Wescott A, Oliver EC. Geometry Effects when Controlling Residual Stresses in Friction Stir Welds by Mechanical Tensioning. Materials Science Forum. 2006;524-525: 71–76.

[457] Staron P, Koçak M, Williams S, Wescott A. Residual stress in friction stir-welded Al sheets. Physica B: Condensed Matter. 2004;350(1-3): E491–E493.

294 LIST OF PUBLICATIONS

First Author

[1] Ahn, J., Chen, L., Davies, C. M. & Dear, J. P. (2016). Parametric optimisation and microstructural analysis on high power Yb-fibre laser welding of Ti–6Al–4V. Optics and Lasers in Engineering, 86, 156-171. [2] Ahn, J., Chen, L., Davies, C. M., & Dear, J. P. (2014). Parametric optimisation and joint heterogeneity characterization of fibre laser welding of AA 2024-T3. Proceedings of the 67th International Institute of Welding Annual Assembly. [3] Ahn, J., Chen, L., Davies, C. M., & Dear, J. P. (2014). Digital Image Correlation for Determination of Local Constitutive Properties of Fibre Laser Welding Joints in AA 2024- T3. Proceedings of the 16th International Conference on Experimental Mechanics. [4] Ahn, J., Chen, L., Davies, C. M., & Dear, J. P. (2014). Residual Stress Measurements in Fibre Laser Beam Welded Plates of Aluminium Alloy AA 2024-T3. Proceedings of the 16th International Conference on Experimental Mechanics.

Second Author

[1] Wang, X., Ahn, J., Kaboglu, C., Yu, L. & Bamber, B. R. K. (2015). Investigation on failure modes and mechanical properties of CFRP-Ti6Al4V hybrid joints with different interface patterns using digital image correlation. Materials & Design, 101, 188-196. [2] Wang, X., Ahn, J., Kaboglu, C., Yu, L. & Bamber, B. R. K. (2015). Characterisation of composite-titanium alloy hybrid joints using digital image correlation. Composite Structures, 140, 702-711. [3] Wang, X., Ahn, J., Bai, Q., Lu, W., & Lin, J. (2015). Effect of forming parameters on electron beam Surfi- Sculpt protrusion for Ti–6Al–4V. Materials & Design, 76, 202-206. [4] Davies, C. M., Ahn, J., Tsunori, M., Dye, D., & Nikbin, K. M. (2015). The Influence of Pre- existing Deformation on GMA Welding Distortion in Thin Steel Plates. Journal of Materials Engineering and Performance, 24(1), 261-273. [5] Wang, X., Ahn, J., Bamber, B. R. K., Mao, Z. & Li. K. (2015). Investigation of failure modes and mechanical properties of hybrid joints of different interface patterns using digital image correlation. Proceedings of the 18th International Conference on Composite Structures.

295 APPENDICES

A. Fibre Laser Welding Parameters Utilised

Table A.1. AA 2024-T3 Focal Sample Laser power Speed Wire feed rate Shielding position number (kW) (m/min) (m/min) Gas (mm) W1 4.9 3.0 0 - W2 4.9 3.0 0 - W3 4.9 2.0 0 - W4 4.9 2.0 0 - W5 4.9 4.0 0 - W6 4.9 5.0 0 - W7 4.9 6.0 0 - Ar W8 4.9 6.0 -4 - W9 4.9 3.0 -2 - W10 4.9 3.0 -4 - W11 4.9 3.0 +2 - W12 4.9 3.0 +4 - W13 3.9 2.4 +4 - W14 2.9 1.8 +4 - W15 4.9 3.0 +4 - W16 4.9 4.0 +4 - W17 4.9 5.0 +4 - W18 4.9 3.0 +2 - W19 4.9 3.0 0 - W20 4.9 3.0 -2 - W21 4.9 3.0 -4 - W22 4.9 3.0 +4 1.5 He W23 4.9 3.0 +4 2.0 W24 4.9 3.0 +4 2.5 W25 4.9 3.0 +4 3.0 W26 4.9 3.0 +4 4.0 W27 4.9 3.0 +4 5.0 W28 4.9 3.0 +4 7.0 W29 4.9 4.0 +4 5.0 W30 4.9 5.0 +4 5.0 W31 4.9 3.0 +4 5.0 W32 4.9 3.0 +2 5.0 Ar W33 4.9 3.0 +2 5.0 W34 4.9 3.0 +4 5.0

296 W35 4.9 3.0 +4 5.0 W36 4.9 3.0 +4 5.0 W37 4.9 3.0 +4 5.0 W38 4.9 3.0 +2 5.0 W39 4.9 3.0 0 5.0 W40 4.9 3.0 0 5.1 W41 4.9 3.0 +4 5.1 W42 4.9 3.0 +4 5.2 W43 4.9 3.0 +4 5.1 W44 4.9 3.0 +2 5.2 W45 4.9 3.0 -2 5.1 W46 4.9 3.0 -4 4.9 W59 4.9 3.0 0 - W60 3.9 3.0 0 - W61 2.9 3.0 0 - W62 1.9 3.0 0 - W63 1.9 2.0 0 - W64 2.9 2.0 0 - W65 3.9 2.0 0 - W66 4.9 2.0 0 - W67 4.9 4.0 0 - W68 3.9 2.0 +2 - W69 3.9 2.0 +2 - W70 3.9 2.0 +4 - W71 3.9 2.0 -2 - W72 3.9 2.0 -4 - W73 4.9 2.0 +4 - W74 2.9 2.0 +4 - W75 1.9 2.0 +4 - W76 2.9 3.0 +4 - W77 3.9 3.0 +4 - W78 4.9 3.0 +4 - W79 4.9 3.0 0 5.2 W80 4.9 3.0 -2 5.0 W81 4.9 3.0 +2 5.1 W82 4.9 3.0 +4 4.1 W83 4.9 3.0 +4 5.2 W84 4.9 3.0 +4 5 W89 2.9 1.7 +4 - W90 2.9 1.5 +4 - W91 2.9 1.3 +4 -

297 W92 2.9 1.1 +4 - W93 1.9 1.7 +4 - W94 1.9 1.5 +4 - W95 1.9 1.3 +4 - W96 1.9 1.0 +4 - W97 2.9 1.5 +4 2.1

Table A.2. Ti-6Al-4V Sample Laser power Speed Focal position Top width Bottom width Undercut Rw number (kW) (m/min) (mm) (mm) (mm) (mm) TH-1-1 1.5 1.8 4.0 2.23 1.33 0.60 0.03 TH-1-4 2.0 1.8 4.0 2.77 2.51 0.91 0.03 TH-1-5 2.0 2.4 4.0 2.07 1.14 0.55 0.05 TH-1-6 2.0 3.0 4.0 2.28 1.76 0.77 0.04 TH-1-7 2.0 3.6 4.0 1.84 0.70 0.38 0.02 TH-1-8 2.0 4.2 4.0 1.76 0.21 0.12 0.01 TH-1-9 2.5 1.8 4.0 3.38 2.90 0.85 0.12 TH-1-10 2.5 2.4 4.0 2.66 2.19 0.82 0.08 TH-1-11 2.5 3.0 4.0 2.24 1.60 0.71 0.07 TH-1-12 2.5 3.6 4.0 2.03 1.10 0.54 0.04 TH-1-13 2.5 4.2 4.0 1.89 0.70 0.37 0.05 TH-1-14 2.5 4.8 4.0 1.78 0.40 0.22 0.02 TH-2-1 3.0 1.8 4.0 2.53 1.92 0.76 0.08 TH-2-2 3.0 2.4 4.0 2.41 1.89 0.78 0.10 TH-2-3 3.0 3.0 4.0 1.96 1.30 0.66 0.07 TH-2-4 3.0 3.6 4.0 2.05 1.32 0.64 0.07 TH-2-5 3.0 4.2 4.0 1.89 1.08 0.57 0.06 TH-2-6 3.0 4.8 4.0 1.95 1.11 0.57 0.07 TH-2-7 3.0 5.4 4.0 1.73 1.10 0.64 0.08 TH-2-8 3.0 6.0 4.0 1.73 0.87 0.50 0.07 TH-2-9 3.5 2.4 4.0 2.26 1.86 0.82 0.08 TH-2-10 3.5 3.6 4.0 1.96 1.20 0.61 0.10 TH-2-11 3.5 4.8 4.0 1.78 1.10 0.62 0.06 TH-2-12 3.5 6.0 4.0 1.77 0.99 0.56 0.08 TH-2-13 3.5 7.2 4.0 1.82 1.00 0.55 0.08 TH-4-1 4.5 2.1 -2.0 4.15 0.00 0.00 0.04 TH-4-2 4.2 2.1 -2.0 3.44 3.24 0.94 0.08 TH-4-4 3.6 2.1 4.0 2.64 1.82 0.69 0.12 TH-4-5 3.3 2.1 4.0 2.77 2.04 0.74 0.15 TH-4-6 3.0 2.1 4.0 2.77 2.08 0.75 0.14 TH-4-7 2.7 2.1 4.0 2.78 2.12 0.76 0.14

298 TH-4-8 2.4 2.1 4.0 2.94 2.30 0.78 0.18 TH-4-9 2.1 2.1 4.0 2.80 2.35 0.84 0.12 TH-4-10 1.8 2.1 4.0 2.49 2.03 0.82 0.06 TH-4-11 1.5 2.1 4.0 2.18 1.45 0.67 0.04 TH-4-12 1.2 2.1 4.0 1.99 0.00 0.00 0.00 TH-5-2 4.2 3.0 4.0 2.53 1.89 0.75 0.12 TH-6-1 4.5 3.9 4.0 2.09 1.37 0.66 0.07 TH-6-2 4.2 3.9 4.0 2.09 1.33 0.64 0.08 TH-6-3 3.9 3.9 4.0 2.15 1.37 0.64 0.08 TH-6-4 3.6 3.9 4.0 2.14 1.28 0.60 0.08 TH-6-5 3.3 3.9 4.0 2.12 1.35 0.64 0.05 TH-6-6 3.0 3.9 4.0 2.00 1.17 0.59 0.06 TH-6-7 2.7 3.9 4.0 1.92 0.91 0.47 0.07 TH-6-8 2.4 3.9 4.0 1.79 0.59 0.33 0.01 TH-6-9 2.1 3.9 4.0 1.60 0.00 0.00 0.02 TH-6-10 1.8 3.9 4.0 1.50 0.00 0.00 0.04 TH-7-1 4.5 6.0 4.0 1.81 1.12 0.62 0.07 TH-7-2 4.2 6.0 4.0 1.80 1.11 0.62 0.06 TH-7-3 3.9 6.0 4.0 1.77 1.12 0.63 0.06 TH-7-4 3.6 6.0 4.0 1.82 1.10 0.60 0.05 TH-7-5 3.4 6.0 4.0 1.91 1.09 0.57 0.06 TH-7-6 3.0 6.0 4.0 1.79 0.86 0.48 0.04 TH-7-7 2.7 6.0 4.0 1.64 0.60 0.37 0.06 TH-7-8 2.4 6.0 4.0 1.69 0.06 0.04 0.03 TH-7-9 2.1 6.0 4.0 1.50 0.00 0.00 0.04 TH-8-2 2.5 3.0 0.0 1.53 1.10 0.72 0.06 TH-8-3 2.5 3.0 4.0 2.20 1.22 0.55 0.07 TH-8-4 2.5 3.0 2.0 1.85 1.46 0.79 0.10 TH-8-5 2.5 3.0 5.5 2.24 1.23 0.55 0.04 TH-8-6 2.5 3.0 -2.0 1.92 1.44 0.75 0.08 TH-8-7 2.5 3.0 -4.0 2.59 1.69 0.65 0.10 TH-8-8 2.5 3.0 -5.5 2.24 1.44 0.64 0.08

299 B. Miscellaneous Results B.1. Micro-indentation Hardness Measurements

a) b) 200 200 W93 ~ W96 W92 P= 1.9 kW, f = +4 mm, Ar P= 2.9 kW, V = 1.1 m/min, 175 V = 1.0 ~ 1.7 m/min 175 f = +4 mm, Ar 1.0 mm

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) 1.0 m/min 75 Top 1.3 m/min 75 1.5 m/min Mid 1.7 m/min Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from centre (mm) Distance from centre (mm)

c) d) 200 200 W91 W90 P= 2.9 kW, V = 1.3 m/min, P= 2.9 kW, V = 1.5 m/min, 175 f = +4 mm, Ar 175 f = +4 mm, Ar

1.0 mm 1.0 mm

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from centre (mm) Distance from centre (mm)

e) f) 200 200 W89 W15 P= 2.9 kW, V = 1.7 m/min, P= 4.9 kW, V = 3.0 m/min, 175 f = +4 mm, Ar 175 f = +4 mm

1.0 mm

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from centre (mm) Distance from weld centre (mm)

g) h) 200 200 W16 W17 P= 4.9 kW, V = 4.0 m/min, P= 4.9 kW, V = 5.0 m/min, f = +4 mm 175 175 f = +4 mm

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker's hardness hardness Vicker's(HV0.1) Top Top 75 Mid 75 Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

Figure B.1 Effect of welding speed on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds

300 a) b) 200 200 W17 W22 P= 4.9 kW, V = 5.0 m/min, P= 4.9 kW, V= 3.0 175 f = +4 mm 175 f = +4 mm, w = 1.5 m/min

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker's hardness hardness Vicker's(HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

c) d) 200 200 W23 W24 P= 4.9 kW, V= 3.0 P= 4.9 kW, V= 3.0 175 f = +4 mm, w = 2.0 m/min 175 f = +4 mm, w = 2.5 m/min

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

e) f) 200 200 W25 W26 P= 4.9 kW, V= 3.0 P= 4.9 kW, V= 3.0 175 f = +4 mm, w = 3.0 m/min 175 f = +4 mm, w = 4.0 m/min

150 150

125 125

100 100

Vicker's hardness hardness Vicker's(HV0.1) Vicker'shardness (HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

g) h) 200 200 W27 W28 P= 4.9 kW, V= 3.0 P= 4.9 kW, V= 3.0 175 f = +4 mm, w = 5.0 m/min 175 f = +4 mm, w = 7.0 m/min

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

Figure B.2 Effect of wire feed rate on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds

301 a) b) 200 200 W72 W71 P= 3.9 kW, V = 2.0 m/min, P= 3.9 kW, V = 2.0 m/min, 175 f = -4 mm, Ar 175 f = -2 mm, Ar

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) Top Top 75 75 Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from centre (mm) Distance from centre (mm) c) d) 200 200 W65 W69 P= 3.9 kW, V = 2.0 m/min, P= 3.9 kW, V = 2.0 m/min, 175 f = 0 mm, Ar 175 f = +2 mm, Ar

150 150

125 125

100 100

Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) Top 75 75 Top Mid Mid Bot Bot 50 50 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from centre (mm) Distance from centre (mm) e) 200 W70 P= 3.9 kW, V = 2.0 m/min, 175 f = +4 mm, Ar

150

125

100 Vicker'shardness (HV0.1) Top 75 Mid Bot 50 -3 -2 -1 0 1 2 3 Distance from centre (mm)

Figure B.3 Effect of focal distance on the micro-indentation hardness distributions of fibre laser welded AA 2024-T3 welds

302 a) b) 450 450 P= 2.0 kW, V = 1.8 m/min P= 2.5 kW, V = 3.0 m/min f= +4 mm, Ar f= +4 mm, Ar

400 400

350 350 Vicker'shardness (HV0.1) 300 Vicker'shardness (HV0.1) 300

Top Top Bot Bot 250 250 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm) c) d)

450 450 P= 3.0 kW, V = 3.6 m/min P= 3.5 kW, V =4.8 m/min f= +4 mm, Ar f= +4 mm, Ar

400 400

350 350 Vicker's hardness hardness Vicker's (HV0.1) 300 Vicker'shardness (HV0.1) 300

Top Top Bot Bot 250 250 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm) e) f) 450 450 P= 3.5 kW, V = 7.2 m/min P= 3.9 kW, V = 3.9 m/min f= +4 mm, Ar f= +4 mm, Ar

400 400

350 350 Vicker's hardness Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) 300 300

Top Top Bot Bot 250 250 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

303 g) h) 450 450 P= 4.2 kW, V = 2.1 m/min P= 4.2 kW, V = 3.0 m/min f= +4 mm, Ar f= +4 mm, Ar

400 400

350 350 Vicker'shardness (HV0.1) Vicker'shardness (HV0.1) 300 300

Top Top Bot Bot 250 250 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm) i) j)

450 450 P= 4.5 kW, V = 3.9 m/min P= 4.5 kW, V = 6.0 m/min f= +4 mm, Ar f= +4 mm, Ar

400 400

350 350 Vicker's hardness Vicker'shardness (HV0.1) 300 Vicker'shardness (HV0.1) 300

Top Top Bot Bot 250 250 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Distance from weld centre (mm) Distance from weld centre (mm)

Figure B.4 Micro-indentation hardness distributions of fibre laser welded Ti-6Al-4V autogenous welds

304 B.2. DIC full field strain measurement during micro-tensile testing

a) V = 3.3 m/min

350 8 t=25% 300 7 t=50% t=75% 6 250 t=90%

5 t=95% 200 t=99% 4

150 Strain Strain (%)

3 Truestress (MPa) 100 FZ 2 HAZ 50 1 Global 0 0 0.00 0.02 0.04 0.06 0.08 0.10 -3 -2 -1 0 1 2 3 True strain Distance from weld centre (mm)

b) V = 3.9 m/min

350 8 t=25% 300 7 t=50% t=75% 6 250 t=90%

5 t=95% 200 t=99% 4

150 Strain Strain (%)

3 Truestress (MPa) 100 FZ 2 HAZ 50 1 Global 0 0 0.00 0.02 0.04 0.06 0.08 0.10 -3 -2 -1 0 1 2 3 True strain Distance from weld centre (mm)

c) V = 4.5 m/min

305 c) V = 4.5 m/min

350 8 t=25% 300 7 t=50% t=75% 6 250 t=90%

5 t=95% 200 t=99% 4

150 Strain Strain (%)

3 Truestress (MPa) 100 FZ 2 HAZ 50 1 Global 0 0 0.00 0.02 0.04 0.06 0.08 0.10 -3 -2 -1 0 1 2 3 True strain Distance from weld centre (mm)

d) V = 5.1 m/min

350 8 t=25% 300 7 t=50% t=75% 6 250 t=90%

5 t=95% 200 t=99% 4

150 Strain Strain (%)

3 Truestress (MPa) 100 FZ 2 HAZ 50 1 Global 0 0 0.00 0.02 0.04 0.06 0.08 0.10 -3 -2 -1 0 1 2 3 True strain Distance from weld centre (mm) Figure B.5 Total strain field of fibre laser welded AA 2024-T3 micro-tensile specimens as a function of welding speed just before fracture at 50x magnification, local and global constitutive data determined for the various weld regions, and development of strain distribution across the weld during load to final fracture at different percentage of the fracture time as determined by DIC

306 a) b) 1200 1200 V=2.1 m/min, f=+2 mm, Ar V=3.0 m/min, f=+2 mm, Ar

1000 1000

800 800

600 600 Truestress (MPa) Truestress (MPa) 400 400 P=1.4 kW P=1.6 kW 2.0 kW 200 200 2.2 kW P=1.8 kW 2.4 kW P=2.0 kW 2.6 kW 0 0 0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10 True strain True strain c) d) 1200 1200 V=5.0 m/min, f=+2 mm, Ar V=6.0 m/min, f=+2 mm, Ar

1000 1000

800 800

600 600 Truestress (MPa) 400 Truestress (MPa) 400 P=1.8 kW P=2.0 kW 3.3 kW 3.6 kW 200 P=2.2 kW 200 3.9 kW 4.2 kW P=2.4 kW 4.5 kW 0 0 0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10 True strain True strain e) f) 1200 1200 P= 2.5 kW, f= +2 mm, Ar P=2.1 kW, V=2.1 m/min, Ar

1000 1000

800 800

600 600 Truestress (MPa) Truestress (MPa) 400 400 2.1 m/min f=0 mm 2.4 m/min f=+2 mm 200 2.7 m/min 200 f=+4 mm 3.0 m/min f=-2 mm 3.3 m/min f=-4 mm 0 0 0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10 True strain True strain

Figure B.6 Global tensile properties of Ti-6Al-4V fibre laser welds processed under different welding conditions of changing a)-d) laser power, e) welding speed and f) focal position

307 B.3. Detailed Design of Tensile Test Specimen X

Figure B.7 Specimen geometry for testing the adhesion of the stiffener to the skin

Figure B.8 Fillet welded AA 2024-T3 T-joint specimen geometries for testing a) with the weld line perpendicular and b) parallel to the tensile y axis

308 a)

Figure B.9 Fillet welded Ti-6Al-4V T-joint specimen geometries for testing a) with the weld line perpendicular and b) parallel to the tensile y axis

309 B.4 Welding Residual Stress Simulation Results

a) b) 250 250

200 200

150 150

100 S11 (Clamp) 100 S11 (Fix)

S22 (Clamp) S22 (Fix)

Stress(MPa) Stress(MPa) 50 S33 (Clamp) 50 S33 (Fix)

0 0

-50 -50 -200 -150 -100 -50 0 50 100 150 200 -200 -150 -100 -50 0 50 100 150 200 Distance from weld centre (mm) Distance from weld centre (mm)

c) d) 250 30

200 20

150 10

100 S11 (Free) 0

S22 (Free) Stress (MPa)

50 S33 (Free) -10 Transverse(MPa) stress Clamp 0 -20 Fix Free -50 -30 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

e) f) 250 30

200 20

150 10

Clamp 100 Fix 0 Free

50 -10

Normalstress (MPa) Longitudinal Longitudinal stress(MPa) Clamp 0 -20 Fix Free -50 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

Figure B.10 Residual stress distributions across the entire width and around the weld in x, y, and z directions under three different mechanical boundary conditions representing clamping fixtures, perfect constraints around the workpiece edges and welding without any fixtures (AA 2024-T3, butt joint)

310 a) b)

250 250

200 200

150 150

100 S11 100 S11

S22 S22

Stress (MPa) Stress (MPa) S33 S33 50 50

0 0

W47 P=4.9 kW, V=3.0 m/min, Ar W99 P=2.9 kW, V=1.5 m/min, Ar -50 -50 -200 -150 -100 -50 0 50 100 150 200 -200 -150 -100 -50 0 50 100 150 200 Distance from weld centre (mm) Distance from weld centre (mm)

c) d) 30 250

20 200

150 10

100 S11 0

S22 Stress (MPa) S33

50 -10 Transverse stress(MPa) W47 -20 0 W99 W102 W102 P=4.9 kW, V=3.0 m/min, Ar, Vf=5.2 m/min -50 -30 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

e) f) 30 250

20 200

150 10 W47 100 W99 0 W102

50 -10

Normal stress(MPa) Longitudinal Longitudinal stress (MPa) W47 0 -20 W99 W102 -50 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

Figure B.11 Residual stress distributions across the entire width and around the weld in x, y, and z directions under three different sets of welding parameters (AA 2024-T3, butt joint)

311 a) b) 60 60 S11 (0% YS) S11 (0% YS) 50 S11 (5% YS) 50 S11 (5% YS) S11 (10% YS) S11 (10% YS) 40 S11 (20% YS) 40 S11 (20% YS) S11 (30% YS) S11 (30% YS) S11 (50% YS) S11 (50% YS) 30 S11 (90% YS) 30 S11 (90% YS) 20 20

10 10

0 0

-10 -10

Transverse(MPa) stress Transverse(MPa) stress -20 -20

-30 -30

-40 -40 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

c) d) 800 800 S22 (0% YS) S22 (0% YS) 700 S22 (5% YS) 700 S22 (5% YS) S22 (10% YS) S22 (10% YS) S22 (20% YS) S22 (20% YS) 600 S22 (30% YS) 600 S22 (30% YS) S22 (50% YS) S22 (50% YS) S22 (90% YS) S22 (90% YS) 500 500

400 400

300 300

Stress (MPa) Stress (MPa) 200 200

100 100

0 0

-100 -100 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

e) f) 30 30 S33 (0% YS) S33 (0% YS) 25 S33 (5% YS) 25 S33 (5% YS) S33 (10% YS) S33 (10% YS) 20 S33 (20% YS) 20 S33 (20% YS) S33 (30% YS) S33 (30% YS) 15 S33 (50% YS) 15 S33 (50% YS) S33 (90% YS) S33 (90% YS) 10 10

5 5

0 0 Normal stress(MPa) Normal stress(MPa) -5 -5

-10 -10

-15 -15

-20 -20 -200 -150 -100 -50 0 50 100 150 200 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from weld centre (mm) Distance from weld centre (mm)

Figure B.12 Residual stress distributions in the a), b) x direction; c), d) y direction; e), f) z direction as a function of the applied tensioning level, pre-loaded to 5, 10, 20 ,30, 50 and 90% of the room temperature yield sterngth of the base metal (Ti-6Al-4V, butt joint)

312 C. User Defined Subroutines used in Finite Element Modelling SUBROUTINE DFLUX(FLUX,SOL,KSTEP,KINC,TIME,NOEL,NPT,COORDS, 1 JLTYP,TEMP,PRESS,SNAME) C------CONICAL GAUSSIAN HEAT SOURCE------C By JOSEPH AHN C AVIC BAMTRI & IMPERIAL COLLEGE C 2013.10.26 C------C SUBROUTINE TO COMPUTE NON-UNIFORM BODY FLUX DISTRIBUTION c AT WELD BEAD DURING TRANSIENT THERMAL ANALYSIS C------3D CONICAL GAUSSIAN HEAT SOURCE TC4 ------C FLUX(1)=MAGNITUDE OF FLUX AT THIS POINT C TEMP=ESTIMATED SURFACE TEMP. C KSTEP=STEP NO. C KINC=INCREMENT NO. C TIME=TIME C NOEL=ELEMENT NO. C NPT=SURFACE INTEGRATION POINT NO. C COORDS=COORDS OF CURRENT POINT C C RBAR=BEAD WIDTH/2=B/2 C X=DIST. OF GAUSS POINT FROM WELD CENTRE C TWOT=HEAT INPUT TIME PERIOD C XI,YI,ZI ARE THE START POINT OF HEAT INPUT C X,Y,Z ARE CURRENT COORDINATES OF THE GAUSS POINT C Q=HEAT INPUT J/s C V=HEAT SOURCE MOVING SPEED mm/s C Qg=CALCULATED BODY FLUX C------C INCLUDE 'ABA_PARAM.INC' C DIMENSION FLUX(2),TIME(2),COORDS(3) DIMENSION XVAL(100),YVAL(100),QY(100),QA(100) CHARACTER*80 SNAME C include 'welding-condition.for' C C-----DEFINE CURRENT COORDINATES C X=COORDS(1) Y=COORDS(2) Z=COORDS(3) C C-----INITIALISE VARIABLES C------ZEE: Top Height C------ZII: Bottom Height C------RO: Average radius C------H: Plate thickness ZEE=2. ZII=0. H=ZEE-ZII REE=0.40 RII=0.30 EE=2.7182818D0

313 ZI=2. YI=0. XI=0. C Q=eff(k)*ec(k)*ev(k) C V=speed Q=3000.*0.80 V=3000./60 PI = 4*atan(1.0) C C-----TO DECIDE THE TOTAL TIME AFTER CURRENT STEP C C IF(TIME(2).GE.0.AND.TIME(2).LE.1)TWTOT=0 C IF(TIME(2).GT.1)TWTOT=TIME(2)-1 TWTOT = time(1) C-----CALCULATE BODY FLUX (Qg) FOR CURRENT GAUSS POINT C RO = REE-(REE-RII)*(ZEE-Z)/H R = ((X-XI)**2+(Y-YI-(V*TWTOT))**2)**0.5 A1 = 9*Q*(EE)**3 A2 = PI*((EE**3)-1)*H*((REE**2)+REE*RII+(RII**2)) A3 = A1/A2 A4 = (-3*(R**2))/(RO**2) Qg = A3*(EE**(A4)) FLUX(1) = Qg C RETURN END

SUBROUTINE USDFLD(FIELD,STATEV,PNEWDT,DIRECT,T,CELENT, 1 TIME,DTIME,CMNAME,ORNAME,NFIELD,NSTATV,NOEL,NPT,LAYER, 2 KSPT,KSTEP,KINC,NDI,NSHR,COORD,JMAC,JMATYP,MATLAYO,LACCFLA) C INCLUDE 'ABA_PARAM.INC' C CHARACTER*80 CMNAME,ORNAME CHARACTER*3 FLGRAY(15) DIMENSION FIELD(NFIELD),STATEV(NSTATV),DIRECT(3,3), T(3,3),TIME(2) C------Ti-6Al-4V Phase Transformation during Welding------C By JOSEPH AHN C AVIC BAMTRI & IMPERIAL COLLEGE C 2014.10.21 C------!Note if model is changed to ask for >15 outputs it will break! !Reassign here appropriately, sub ref guide sec 2.1.6 DIMENSION ARRAY(15),JARRAY(15),JMAC(*),JMATYP(*),COORD(*) C ------C USFLD routine to calculate change due to phase transformation. C Solution dependant state variables - must be defined and initialised C in the ABAQUS input file. Use Mechanical Model. C STATEV(1) Current temp stored as previous temp C STATEV(2) Time increment DTIME C STATEV(3) Heating or Cooling rate dT/dt C STATEV(4) Update Field(1) to store max Temp C STATEV(5) Beta Phase C STATEV(6) Alpha Phase

314 C STATEV(7) Martensite Phase C STATEV(8) Old Martensite Phase C STATEV(9) Old Beta Phase C STATEV(10) Old Alpha Phase C FIELD(1) Current Temp C FIELD(2) DTIME C FIELD(3) DT/Dt C FIELD(4) Connect SDV7 and SDV8 C ------C USER CODE START C Reading instantaneous temperature in direction 11 !grabs the temperatures from the solution (NT does not work here) CALL GETVRM('TEMP',ARRAY,JARRAY,FLGRAY,JRCD,JMAC,JMATYP,MATLAYO, 1 LACCFLA) FIELD(1)=ARRAY(1) !temp as field variable DTEMP=FIELD(1)-STATEV(1) !dT Current temp - previous stored temp FIELD(2)=DTIME !dt FIELD(3)=DTEMP/FIELD(2) !dT/dt

! test result of GETVRM say goodbye to Abaqus if something went wrong if (jrcd == 1) then ! write error to dat file write(6,*) "GETVRM routine error!!!" ! stop Abaqus call XIT endif ! a-h,o-z are real and i-n are integers ! define JMA coefficients k and n an=2.5D0 !(n) xc=734.59204D0 y=3.84D-05 H=0.01375D0 wi=58.38677D0 we=35.81163D0 EE=2.7182818D0

IF (FIELD(1).LT.xc) THEN ak=y+(H*EE**(-0.5*(FIELD(1)-xc)**2.0/(wi**2.0))) ako=y+(H*EE**(-0.5*(STATEV(1)-xc)**2.0/(wi**2.0))) ELSE ak=y+(H*EE**(-0.5*(FIELD(1)-xc)**2.0/(we**2.0))) ako=y+(H*EE**(-0.5*(STATEV(1)-xc)**2.0/(we**2.0))) ENDIF

C Additivity rule k = k1 (new) and ko = k0 (old) C FAO = alpha fraction at t0 (old) C FAN = alpha fraction at t1 (new) / FBN = beta fraction at t1 (new) C TF = additional step added at t1 (t1f) C OAE = equilibrium alpha fraction at t0 (old) C FAEQ and FBEQ are equilibrium fractions at t1 (new, current) C FAM and FAD are alpha`+alpham; martensite fractions at t1 (new, current)

IF (FIELD(1).GE.1650.0D0) THEN FBEQ = 1.00D0 ! just assume equals to 1, prevent overshoot ELSEIF ((FIELD(1).GE.980.0D0).AND.(FIELD(1).LT.1650.0D0)) THEN

315 FBEQ = 1.00D0 ELSEIF (FIELD(1).LE.20.0D0) THEN FBEQ = 0.07526D0 ELSE FBEQ = (0.92*EE**((980.0D0-FIELD(1))*-0.0085))+0.075 ENDIF C IF (FIELD(1).GE.980.0D0) THEN FAEQ = 0.00D0 ELSEIF (FIELD(1).LE.20.0D0) THEN FAEQ = 0.92474D0 ELSE FAEQ = 1.0D0-((0.92*EE**((980.0D0-FIELD(1))*-0.0085))+0.075) ENDIF C OAE = 1.0-((0.92*EE**((980.0-STATEV(1))*-0.0085))+0.075) FAO = (1.0-EE**(-ako*(DTIME**2.5)))*OAE TF = ((-log(1.0-(FAO/FAEQ)))/ak)**(1.0/an) FAM = (1.0-EE**(-0.015*(650-FIELD(1))))*(1.0-FBEQ) FAD = 1.0-EE**(-0.015*(650.0-FIELD(1))) C IF (FIELD(1).GT.650.0D0) THEN FAD = 0.00D0 FAM = 0.00D0 ELSEIF (FIELD(1).LE.25.0D0) THEN FAD = 1.00D0 !tends to 1.0 at low T FAM = 0.925D0 !tends to 0.925 at low T ELSE FAM = (1.0-EE**(-0.015*(650.0-FIELD(1))))*(1.0-FBEQ) FAD = 1.0-EE**(-0.015*(650.0-FIELD(1))) ENDIF C C======C Store the current temp as a solution dependent state (previous, old) STATEV(1)=FIELD(1) STATEV(2)=FIELD(2) STATEV(3)=FIELD(3) C======C Initial phase fraction before welding step (kstep 3) IF (KSTEP.LT.3) THEN STATEV(4)=0.00D0 STATEV(5)=0.07526D0 STATEV(6)=0.92474D0 STATEV(7)=0.00D0 STATEV(8)=0.00D0 ENDIF C======C----- LIQUID define variable as 0=solid 1=liquid 2=vapour C FIXED PROBLEM BETWEEN MELT AND SOLID to GE & LE C ! If the material is still solid C IF (STATEV(4).EQ.0) THEN !Check if the temperature is over liquidus 1 = melt C IF (FIELD(1).GT.1650.0) THEN C STATEV(4)=1.0 !If the temp is below melting remain as solid phase

316 C ELSEIF ((FIELD(1).GE.1604.0) .AND. (FIELD(1).LE.1650.0)) THEN C STATEV(4)=(FIELD(1)-1604.0)/(1650.0-1604.0) C ELSE C STATEV(4)=0.0 C ENDIF C IF (FIELD(1).LT.STATEV(4)) THEN STATEV(4)=STATEV(4) ELSE STATEV(4)=FIELD(1) ENDIF C C======CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC CCCCCCCCCCCCCCCC IF ((KSTEP.GE.3) .AND. (KSTEP.LT.5)) THEN ! before reheat C step 5 equals to clamp free state in stress model C======C-----Martensite define variable as 0=0% ALPHA` 1=100% ALPHA` C Working Fine! ! If the material is 0% ALPHA` C IF (STATEV(6).EQ.0) THEN

!Check if the temperature is over 980 beta transus IF (FIELD(1).GE.980.00) THEN STATEV(7)=0.00 !If the temp is below beta transus and heating dT/dt >0 ELSEIF ((FIELD(1).LT.980.00) .AND. (FIELD(3).GT.0.00)) THEN STATEV(7)=0.00 ! equilibrium phase fraction of alpha ! If the temp is below beta transus and cooling dT/dt<0 ELSEIF ((FIELD(1).LT.980.00) .AND. (FIELD(3).LE.0.00)) THEN ! Martensite + massive alpha between 410>dT/dt>25 IF ((FIELD(3).LE.-25.00) .AND. (FIELD(3).GT.-410.00)) THEN IF (STATEV(4).GE.1200.00) THEN !limit Tmax for trans start STATEV(7)=FAM ELSE STATEV(7)=0.00 ENDIF !dT/dt >= to 410 martensite and below Ms ELSEIF ((FIELD(3).LE.-410.00).AND.(FIELD(1).LE.650.00)) THEN IF (STATEV(4).GE.1200.00) THEN !limit Tmax for trans start STATEV(7)=FAM ! seems always zero since dTdt not fast ELSE ! FAD above causing sudden jump to 0.99 so use FAM STATEV(7)=0.00 ENDIF !dT/dt >= to 410 martensite and above Ms ELSEIF ((FIELD(3).LE.-410.00).AND.(FIELD(1).GT.650.00)) THEN STATEV(7)=0.00 ELSE !dT/dt < 25 STATEV(7)=0.00! phase fraction of beta non-isothermal ENDIF ELSE STATEV(7)=0.00 ! no change, base metal beta=0.1 ENDIF C Limit max and min phase fractions overshoot (kstep 3)

317 IF (STATEV(7) .GE. 1.00) THEN STATEV(7)=1.00 ELSEIF (STATEV(7) .LE. 0.00) THEN STATEV(7)=0.00 ENDIF C STATEV(8) is the transformed martensite phase(previous) C Keep the transformed martensite fraction and prevent backwards trans. IF (STATEV(8) .GE. 1.00) THEN STATEV(8)=1.00 ELSEIF (STATEV(8) .LE. 0.00) THEN STATEV(8)=0.00 ENDIF C DIFF=STATEV(7)-STATEV(8) IF (DIFF.LT.0.00) THEN STATEV(7)=STATEV(8) ENDIF

C ! Initialise field 4 in abaqus inp file. FIELD(4)=STATEV(7) STATEV(8)=FIELD(4) C C======C-----BETA define variable as 0=0% BETA 1=100% BETA C ! If the material is 0% BETA !FAEQ=1-FBEQ C IF (STATEV(5).EQ.0) THEN C IF (STATEV(5).NE.0.0) THEN ! when beta frac=0 dont calc again ! Check if the temperature is over 1650 IF (FIELD(1).GE.1650.00) THEN STATEV(5)=FBEQ ! If the temp is below melting and above beta transus ELSEIF ((FIELD(1).GE.980.00) .AND. (FIELD(1).LT.1650.00)) THEN STATEV(5)=FBEQ ! If the temp is below beta transus and heating dT/dt>0 ELSEIF ((FIELD(1).LT.980.00) .AND. (FIELD(3).GT.0.00)) THEN STATEV(5)=FBEQ ! equilibrium phase fraction of beta ! If the temp is below beta transus and cooling dT/dt<0 ELSEIF ((FIELD(1).LT.980.00) .AND. (FIELD(3).LE.0.00)) THEN ! Martensite + massive alpha between 410>dT/dt>25 IF ((FIELD(3).LE.-25.00) .AND. (FIELD(3).GT.-410.00)) THEN IF (STATEV(4).GE.1200.00) THEN !limit Tmax for trans start STATEV(5)=1.00-FAD ! Beta goes to zero ELSE STATEV(5)=FBEQ ENDIF !dT/dt >= to 410 martensite and below Ms ELSEIF ((FIELD(3).LE.-410.00).AND.(FIELD(1).LE.650.00)) THEN IF (STATEV(4).GE.1200.00) THEN !limit Tmax for trans start STATEV(5)=1.00-FAD ! Beta goes to zero ELSE STATEV(5)=FBEQ ENDIF !dT/dt >= to 410 martensite and above Ms ELSEIF ((FIELD(3).LE.-410.00).AND.(FIELD(1).GT.650.00)) THEN

318 STATEV(5)=1.00 ELSE !dT/dt < 25 STATEV(5)=FBN ! Phase fraction of beta non-isothermal ENDIF ELSE STATEV(5)=FBEQ ! no change, base metal beta=0.1 ENDIF C C If transformed martensite fraction is greater than zero then C the rate of transformation is diffusionless so the beta frac = 0 C below Martensite finish temp Mf (around 400) then beta = 0 IF (STATEV(7).GT.0.00) THEN STATEV(5)=1.00-FAD ! During cooling (not diffusion controlled) ENDIF C STATEV(9)=STATEV(5) C C Limit max and min phase fractions overshoot (kstep 3) IF (STATEV(5).GT.1.00) THEN STATEV(5)=1.00 ELSEIF (STATEV(5).LT.0.00) THEN STATEV(5)=0.00 ENDIF C ENDIF C If transformed alpha fraction is zero then the remaining fractions C only consist of either beta or martensite (+massive) phases C The maintained fraction of beta phase depends on the presence of C alpha phase so there are two cases available. C IF (STATEV(6).EQ.0.0) THEN C STATEV(5)=1.0-STATEV(7) C ENDIF C C======C-----ALPHA define variable as 0=0% ALPHA 1=100% ALPHA C ! If the material is 0% ALPHA C IF (STATEV(6).NE.0.0) THEN ! when alpha frac=0 dont calc again !Check if the temperature is over 980 beta transus IF (FIELD(1).GE.980.00) THEN STATEV(6)=0.00 !If the temp is below beta transus and heating dT/dt >0 ELSEIF ((FIELD(1).LT.980.00) .AND. (FIELD(3).GT.0.00)) THEN STATEV(6)=FAEQ ! equilibrium phase fraction of alpha !If the temp is below beta transus and cooling dT/dt <0 ELSEIF ((FIELD(1).LT.980.00) .AND. (FIELD(3).LE.0.00)) THEN ! Martensite + massive alpha between 410>dT/dt>25 IF ((FIELD(3).LE.-25.00) .AND. (FIELD(3).GT.-410.00)) THEN IF (STATEV(4).GE.1200.00) THEN !limit Tmax for trans start STATEV(6)=1.0-FAM-STATEV(5) !FIX HERE make total frac=1 ELSE STATEV(6)=FAEQ ENDIF !dT/dt >= to 410 martensite and below Ms ELSEIF ((FIELD(3).LE.-410.00).AND.(FIELD(1).LE.650.00)) THEN STATEV(6)=0.00

319 !dT/dt >= to 410 martensite and above Ms ELSEIF ((FIELD(3).LE.-410.00).AND.(FIELD(1).GT.650.00)) THEN STATEV(6)=0.00 ELSE !dT/dt < 25 STATEV(6)=FAN ! Phase fraction of beta non-isothermal ENDIF ENDIF C If transformed martensite fraction is greater than zero then C the rate of transformation is diffusionless so the alpha frac = 0 C IF (STATEV(7).GT.0.0) THEN C STATEV(6)=0.0 ! During cooling (not diffusion controlled) C ENDIF C C If transformed beta fraction is zero then the remaining fractions C only consist of either alpha (+massive) or martensite phases C The maintained fraction of alpha phase depends on the presence of C martensite phase so there are two cases available. IF (STATEV(7).GT.0.00) THEN STATEV(6)=1.00-FAM-STATEV(5) ENDIF C C Limit max and min phase fractions overshoot (kstep 3) IF (STATEV(6) .GT. 1.00) THEN STATEV(6)=1.00 ELSEIF (STATEV(6) .LT. 0.00) THEN STATEV(6)=0.00 ENDIF C STATEV(10)=STATEV(6) C C======ENDIF C C DISABLED PWHT PHASE TRANSFORMATION SINCE TEMP LESS THAN C ALPHA DISSOLUTIONN TEMP SO PREVIOUS PHASES RETAINED! C Tdiss=750C C C======C IF (KSTEP.EQ.6) THEN ! PWHT (reheat) step 6 in stress model C Reheat: two conditions (initial composition a` + a mass or alpha+beta) C 1. alpha' to alpha`+ alpha + beta (incomplete alpha` decomposition) C 2. alpha to alpha + beta (equilibrium fractions, simply cycle) C Heating rate 20C/s and cooling also slow (RT cooling so dT/dt<410C/s) C IF (STATEV(8).LE.0.00) THEN C STATEV(5)=FBEQ ! beta C STATEV(6)=FAEQ ! alpha C STATEV(7)=0.00 ! martensite C ELSE C STATEV(7)=STATEV(8)-(0.92467D0-FAM) ! martensite C STATEV(5)=STATEV(9)+((0.92467D0-FAM)/2.00) ! beta C STATEV(6)=STATEV(10)+((0.92467D0-FAM)/2.00) ! alpha C ENDIF C ENDIF C C IF (STATEV(7) .GE. 1.00) THEN

320 C STATEV(7)=1.00 C ELSEIF (STATEV(7) .LE. 0.00) THEN C STATEV(7)=0.00 C ENDIF C IF (STATEV(6) .GT. 1.00) THEN C STATEV(6)=1.00 C ELSEIF (STATEV(6) .LT. 0.00) THEN C STATEV(6)=0.00 C ENDIF C IF (STATEV(5).GT.1.00) THEN C STATEV(5)=1.00 C ELSEIF (STATEV(5).LT.0.00) THEN C STATEV(5)=0.00 C ENDIF C______C IF (KSTEP.EQ.7) THEN ! 2nd cooling step 7 in stress model C Cooling: two conditions (initial composition 100% alpha` or alpha+beta) C Heating rate 20C/s and cooling also slow (RT cooling so dT/dt<410C/s) C No martensite formed during 2nd cooling since cooling rate slow C IF (STATEV(8).LE.0.00) THEN C STATEV(5)=FBEQ ! beta C STATEV(6)=FAEQ ! alpha C STATEV(7)=0.00 ! martensite C ELSE C STATEV(5)=FBEQ*(1-STATEV(7)) ! beta (1-alpha`) normalise C STATEV(6)=FAEQ*(1-STATEV(7)) ! alpha (1-alpha`) normalise C STATEV(7)=0.46886 ! martensite (frac. dep on HT temp) C ENDIF C ENDIF C C IF (STATEV(7) .GE. 1.00) THEN C STATEV(7)=1.00 C ELSEIF (STATEV(7) .LE. 0.00) THEN C STATEV(7)=0.00 C ENDIF C IF (STATEV(6) .GT. 1.00) THEN C STATEV(6)=1.00 C ELSEIF (STATEV(6) .LT. 0.00) THEN C STATEV(6)=0.00 C ENDIF C IF (STATEV(5).GT.1.00) THEN C STATEV(5)=1.00 C ELSEIF (STATEV(5).LT.0.00) THEN C STATEV(5)=0.00 C ENDIF C Heat treatment at 600C so the FAM (a`+amassive) at 600 = 0.46886 C Transformation from beta to alpha on cooling only possible from C the remaining fraction 1-FAM @ 600C =0.53114 so normalise C FAEQ and FBEQ by multiplying by 0.53114 so the sum = 0.53114 C======RETURN END

SUBROUTINE uexpan(expan,dexpandt,temp,time,dtime,predef,dpred, $ statev,cmname,nstatv,noel)

321 c include 'aba_param.inc' c character*80 cmname c dimension expan(*),dexpandt(*),temp(2),time(2),predef(*), $ dpred(*),statev(nstatv) C------Ti-6Al-4V Transformation Strain during Welding------C By JOSEPH AHN C AVIC BAMTRI & IMPERIAL COLLEGE C 2015.01.08 C------C EXPAN(1) : Increment of thermal strain isotropic C TEMP(1) : Current temperature at the end of the increment C TEMP(2) : TEMPERATURE INCREMENT c Tref : strain free reference value *---- Reference temperature Tref = 20.0D0 *---- Set Temperature Tc= temp(1) ! temp at the end of the increment Tp= temp(1) - temp(2) ! temp at the start of the increment c T_melt : melting temperature T_MELT = 1650.0D0 * *---- Calculate linear coefficient of thermal expansion alpha * C Model PWL3 for beta phase C beta phase has bcc structure C Equation double by1, by2; xai= 8.59798D-06 xki= 3.21427D-09 xxi= 702.2944573D0 xkt= 7.317D-08 xxt= 920.9976796D0 xks= -5.49396D-09 xat= -4.05315D-05 xas= 3.19178D-05 by1=xai+(xki*xxi) by2=by1+xkt*(xxt-xxi) C IF( Tp .GE. T_MELT ) THEN pby= 2.28527D-05 ELSEIF(( Tp .GE. xxt ).and.( Tp .LT. T_MELT )) THEN pby=by2+xks*(Tp-xxt) ELSEIF(( Tp .GE. xxi ).and.( Tp .LT. xxt )) THEN pby=by1+xkt*(Tp-xxi) ELSEIF( Tp .LT. 30.0 ) THEN pby=8.69441D-06 ELSE pby=xai+xki*Tp ENDIF

IF( Tc .GE. T_MELT ) THEN cby= 2.28527D-05 ELSEIF(( Tc .GE. xxt ).and.( Tc .LT. T_MELT )) THEN

322 cby=by2+xks*(Tc-xxt) ELSEIF(( Tc .GE. xxi ).and.( Tc .LT. xxt )) THEN cby=by1+xkt*(Tc-xxi) ELSEIF( Tc .LT. 30.0) THEN cby=8.69441D-06 ELSE cby=xai+xki*Tc ENDIF C ------C Model PWL2 for alpha phase (including alpha` and alpha massive) C alpha and alpha` has hcp structure alpha`` and alpha``` not considered C Equation double ay1; yai= 8.82943D-06 yki= 2.95031D-09 yxi= 761.2245859D0 ykt= 1.25752D-08 yat= 1.50277D-06 ay1=yai+(yki*yxi) C IF( Tc .GE. 980.0D0 ) THEN cay= 0.0D0 ELSEIF(( Tc .GE. yxi ).and.( Tc .LT. 980.0D0 )) THEN cay=ay1+ykt*(Tc-yxi) ELSEIF( Tc .LT. 30.0) THEN cay=8.91794D-06 ELSE cay=yai+yki*Tc ENDIF C IF( Tp .GE. 980.0D0 ) THEN pay= 0.0D0 ELSEIF(( Tp .GE. yxi ).and.( Tp .LT. 980.0D0 )) THEN pay=ay1+ykt*(Tp-yxi) ELSEIF( Tp .LT. 30.0 ) THEN pay=8.91794D-06 ELSE pay=yai+yki*Tp ENDIF C ------*---- Calculate thermal strain etha = ABS(STATEV(6))*((cay * ( Tc -Tref ))-(pay * ( Tp -Tref ))) ethm = ABS(STATEV(7))*((cay * ( Tc -Tref ))-(pay * ( Tp -Tref ))) ethb = ABS(STATEV(5))*((cby * ( Tc -Tref ))-(pby * ( Tp -Tref ))) eth = etha + ethb +ethm ! a*dt -> a2*t2 - a1*t1 *---- Calculate transformation strain C Model (alpha to beta) Poly C y = a0+a1*x+a2*x^2+a3*x^3+a4*x^4+a5*x^5+a6*x^6+a7*x^7+a8*x^8+a9*x^9 va0=-0.039295069D0 va1=-7.02D-05 va2=1.20D-06 va3=-1.07D-08 va4=5.21D-11 va5=-1.43D-13 va6=2.23D-16 va7=-1.92D-19

323 va8=8.28D-23 va9=-1.34D-26 C ------etrab1 = va0+va1*Tc+va2*(Tc**2.0)+va3*(Tc**3.0) etrab2 = va4*(Tc**4.0)+va5*(Tc**5.0)+va6*(Tc**6.0) etrab3 = va7*(Tc**7.0)+va8*(Tc**8.0)+va9*(Tc**9.0) etrab = etrab1 + etrab2 + etrab3 etrba = (-1.0)*etrab

IF (Tp.GE.T_MELT) THEN expan(1) = 0.00D0 ELSEIF ((Tp.GE.980.0D0) .and. (Tp.LT.T_MELT)) THEN expan(1) = eth ! etr = 0 ELSEIF (TEMP(2).GT.0.0000D0) THEN ! heating alpha (hcp) to beta (bcc) IF (Tp.LE.980.0D0) THEN ! hcp -> bcc etrab expan(1) = eth + (etrab*abs(STATEV(5)-STATEV(9))) !eth+etr ENDIF ELSEIF (TEMP(2).LT.0.0000D0) THEN ! cooling beta (bcc) to alpha (hcp) IF (Tp.LE.980.0D0) THEN ! bcc -> hcp etrba expan(1) = eth + (etrba*abs(STATEV(5)-STATEV(9))) !eth+etr ENDIF ! etrba for beta to alpha and alpha` ELSE expan(1) = eth ENDIF C RETURN END

324

View publication stats