Development of A Right Angle Friction Stir Welding (RAFSW) Technique To Assemble Aluminum Products

Development of a right angle friction stir welding
(RAFSW) technique to assemble aluminum products
Thèse
Mahboubeh Momeni
Doctorat en génie mécanique

Philosophiæ doctor (Ph. D.)
Québec, Canada
© Mahboubeh Momeni, 2020

Development of a right angle friction stir welding
(RAFSW) technique to assemble aluminum
products
Thèse
Mahboubeh Momeni
Sous la direction de:
Michel Guillot, directeur de recherche

Résumé
Aujourd'hui, le soudage par friction-malaxage (FSW) a attiré beaucoup d'attention dans les secteurs
universitaires et industriels. Malgré ses avantages importants par rapport aux techniques de soudage
par fusion, il n'est pas largement utilisé dans l'industrie actuelle, principalement en raison des coûts
d'équipement élevés et des redevances. Certaines autres raisons sont les forces de processus élevées,
le besoin d'un serrage puissant, le manque de directives concernant la fenêtre de travail efficace des
paramètres du processus et l'effet du traitement thermique après soudage. En outre, il est nécessaire
d'avoir une conception d'outils appropriée pour différentes applications, géométries et configurations
de soudage. Pour surmonter ces problèmes, une technique FSW rentable appelée FSW à angle droit
(RAFSW) a récemment été introduite par l'équipe PI2 / REGAL de l'Université Laval. Il est essentiel
de développer et d'étudier ses différents aspects pour rendre la technique fiable pour une large
utilisation industrielle. Dans cette thèse, l'objectif est de fournir aux utilisateurs potentiels des
directives et des fenêtres de travail efficaces pour les paramètres du processus de soudage à des
vitesses de soudage élevées applicables à différentes configurations et géométries. Une conception
d'outils appropriée pour différentes applications est également un autre aspect à explorer. De plus,
l'effet du traitement thermique après soudage sera étudié. Enfin, la technique sera adaptée pour être
mise en œuvre sur des routeurs CNC de grande taille et à faible coût afin d'assembler de grands
panneaux en aluminium à de faibles forces de soudage sans avoir besoin d'un serrage solide.
iii
Abstract
Today, friction stir welding (FSW) has attracted much attention in both academic and industrial
sectors. In spite of its prominent advantages over fusion welding techniques, it is not widely used in
today's industry mainly due to high equipment costs and royalties. Some other reasons are high
process forces, need for powerful clamping, lack of guidelines regarding efficient working window
of process parameters and the effect of post weld heat treatment. Furthermore, it is needed to have a
proper tool design for different applications, geometries, and welding configurations. To overcome
these issues, a cost-effective FSW technique called FSW at right angle (RAFSW) has been recently
introduced by PI2/REGAL team at Laval University. It is essential to develop and study its different
aspects to make the technique reliable for widespread industrial use. In this thesis, the aim is to
provide potential users with guidelines and efficient working windows for welding process
parameters at high welding speeds applicable for different configurations and geometries. Proper tool
design for different applications is another aspect to be explored, as well. Moreover, the effect of post
weld heat treatment will be studied. Finally, the technique will be adapted to implement on big-size,
low-cost CNC routers to assemble large aluminum panels at low welding forces without the need for
sturdy clamping.
iv
Table of Contents
Résumé ............................................................................................................................................... iii
Abstract .............................................................................................................................................. iv
Table of Contents ................................................................................................................................ v
List of Figures ..................................................................................................................................... x
List of Tables................................................................................................................................... xvii
Dedication ........................................................................................................................................ xix
Acknowledgment .............................................................................................................................. xx
Preface .............................................................................................................................................. xxi
Introduction ......................................................................................................................................... 1
General background ........................................................................................................................ 1
Benefits and challenges ................................................................................................................... 1
Literature Review............................................................................................................. 3
1.1 FSW Process and the joint specifications.................................................................................. 3
1.1.1 Joint Configuration............................................................................................................. 4
1.1.2 Heat input ........................................................................................................................... 4
1.1.3 Microstructural Evolution .................................................................................................. 5
1.1.4 FSW Joints Defects ............................................................................................................ 7
1.1.5 Residual Stresses ................................................................................................................ 8
1.1.6 Hardness ............................................................................................................................. 9
1.1.7 Tensile Properties ............................................................................................................... 9
1.1.8 Fatigue behavior ............................................................................................................... 10
1.2 Welding parameters................................................................................................................. 11
1.3 Tool Design ............................................................................................................................. 11
1.4 FSW machines ........................................................................................................................ 13
1.5 Making the FSW process affordable ....................................................................................... 16
1.6 Study the effect of tool design and process parameters .......................................................... 16
v
1.6.1 Effect of tool tilt angle on generated axial force .............................................................. 16
1.6.2 Effect of Tool Design on Downward Axial Force ........................................................... 18
1.6.3 Effect of Preheating on Axial Force ................................................................................. 19
1.6.4 Effect of plunge depth, tool rotation speed, and transverse speed on generated axial force
................................................................................................................................................... 20
1.7 Effect of tool design and process parameters on mechanical properties of the FSW joints .... 23
1.8 Fatigue behavior of FSW joints made by different process parameters and tool designs ....... 31
1.9 Effect of joint fit-up variations on the quality of the FSW joints ............................................ 32
1.10 Effect of post weld heat treatment (PWHT) on mechanical properties and fatigue behavior of
FSW joints..................................................................................................................................... 33
1.11 Applications of FSW ............................................................................................................. 37
Problems, Objectives, and Methodology ....................................................................... 42
2.1 Problem statement ................................................................................................................... 42
2.1.1 FSW.................................................................................................................................. 42
2.1.2 RAFSW ............................................................................................................................ 42
2.1.3 Tool design and process parameters ................................................................................. 43
2.1.4 PWHT .............................................................................................................................. 44
2.1.5 Applications ..................................................................................................................... 44
2.2 Hypothesis ............................................................................................................................... 46
2.3 Objectives of the thesis ........................................................................................................... 46
2.4 General Methodology.............................................................................................................. 47
Development of Friction Stir Welding Technique at Right Angle (RAFSW) Applied on

Butt-Joint of AA6061-T6 Aluminum Alloy...................................................................................... 49
3.1 Résumé .................................................................................................................................... 50
3.2 Abstract ................................................................................................................................... 50
3.3 Introduction ............................................................................................................................. 51
3.4 Experimental procedure .......................................................................................................... 52
3.4.1 Material properties ........................................................................................................... 53
vi
3.4.2 Quality evaluation ............................................................................................................ 53
3.4.3 Metallography .................................................................................................................. 53
3.4.4 Micro-hardness ................................................................................................................. 54
3.4.5 Post weld heat treatment .................................................................................................. 54
3.5 Design of experiments and modeling ...................................................................................... 54
3.5.1 Design of experiments ...................................................................................................... 54
3.5.2 ANN modeling ................................................................................................................. 57
3.6 Results and discussions ........................................................................................................... 60
3.6.1 Effect of tool plunge depth ............................................................................................... 60
3.6.2 Axial force during RAFSW .............................................................................................. 61
3.6.3 Ultimate tensile strength of RAFSW specimens .............................................................. 62
3.6.4 Robustness of the RAFSW process .................................................................................. 63
3.6.5 Effect of artificial aging ................................................................................................... 65
3.6.6 Macrostructure examinations ........................................................................................... 66
3.6.7 Microstructure examinations ............................................................................................ 66
3.6.8 Micro-hardness examinations........................................................................................... 67
3.7 Conclusion............................................................................................................................... 68
Post Weld Heat Treatment effects on Mechanical Properties and Microstructure of

AA6061-T6 Butt Joints Made by Friction Stir Welding at Right Angle (RAFSW) ......................... 70
4.1 Résumé .................................................................................................................................... 71
4.2 Abstract ................................................................................................................................... 71
4.3 Introduction ............................................................................................................................. 72
4.4 Materials and Methods ............................................................................................................ 73
4.4.1 Preparation of the joints ................................................................................................... 73
4.4.2 Post weld heat treatment .................................................................................................. 73
4.4.3 Characterization ............................................................................................................... 74
4.5 Results and Discussion ............................................................................................................ 75
vii
4.5.1 Natural aging .................................................................................................................... 75
4.5.2 Artificial aging ................................................................................................................. 77
4.5.3 Solubilizing followed by artificial aging (W+T6) ............................................................ 82
4.6 Conclusions ............................................................................................................................. 85
Effect of Tool Design and Process Parameters on Lap Joints Made by Right Angle
Friction Stir Welding (RAFSW) ....................................................................................................... 86
5.1 Résumé .................................................................................................................................... 87
5.2 Abstract ................................................................................................................................... 87
5.3 Introduction ............................................................................................................................. 88
5.4 Materials and Methods ............................................................................................................ 90
5.5 Results and Discussion ............................................................................................................ 92
5.5.1 Design of Experiments ..................................................................................................... 92
5.5.2 Artificial Neural Network Modeling ................................................................................ 95
5.5.3 Effect of Tool Geometry Parameters on Welding Force and Tensile Shear Force .......... 98
5.5.4 Effect of Process Parameters on Welding Force and Tensile Shear Force .................... 101
5.6 Conclusions ........................................................................................................................... 104
Implementation of Right Angle Friction Stir Welding (RAFSW) to Assemble the Side
Panels of Truck Box ........................................................................................................................ 105
6.1 Résumé .................................................................................................................................. 106
6.2 Abstract ................................................................................................................................. 106
6.3 Introduction ........................................................................................................................... 107
6.3.1 Friction stir welding ....................................................................................................... 107
6.3.2 Truck box side panels ..................................................................................................... 108
6.3.3 Objectives of this research ............................................................................................. 108
6.4 Material ................................................................................................................................. 109
6.5 Optimization of lap joint welding ......................................................................................... 109
6.5.1 Apparatus ....................................................................................................................... 109
6.5.2 Design of experiments .................................................................................................... 111
viii
6.5.3 Artificial Neural Network Modeling .............................................................................. 112
6.5.4 Effect of process parameters on the downward axial force and failure force ................ 114
6.6 Adaptation of the RAFSW process ....................................................................................... 116
6.6.1 Adaptation of RAFSW technique and its apparatus ....................................................... 116
6.6.2 Validation of the tool and tool holder ............................................................................ 117
6.7 Application and test of larger panel assemblies .................................................................... 118
6.7.1 CNC router set-up and welding sequence ...................................................................... 118
6.7.2 Strength validation of truck side panels ......................................................................... 119
6.8 Conclusion............................................................................................................................. 123
Conclusions and Perspectives ......................................................................................................... 125
Overview of the project ............................................................................................................... 125
Conclusions ................................................................................................................................. 125
Perspectives ................................................................................................................................. 127
References ....................................................................................................................................... 128
Appendix: Artificial neural network modeling ............................................................................... 135
Background ................................................................................................................................. 135
Theory ......................................................................................................................................... 135
A case study: details of developing ANN models....................................................................... 139
ix
List of Figures
Figure 1: (a) Schematic of FSW in a butt-joint configuration, (b) FSW seam weld, (c) an FSW tool,
which consists of a pin and shoulder [2]. ............................................................................................ 3
Figure 2: FSW joint configurations: (a) square butt, (b) edge butt, (c) T butt, (d) lap butt, (e) multiple
lap, (f) T lap, and (g) fillet joint [1]. .................................................................................................... 4
Figure 3: Heat input vs. welding speed input [6]. ............................................................................... 5
Figure 4: (a) Typical microstructure of friction stir welded aluminum alloys, (b) higher magnification
in the retreating side, and (c) higher magnification in the advancing side [7]. ................................... 6
Figure 5: Precipitate size, distribution, and shape in: (a) base metal, (b) HAZ, (c) TMAZ near HAZ,
and (d) TMAZ near nugget zone [1]. .................................................................................................. 7
Figure 6: FSW common defects and the safe range of tool rotational speed and traverse speed [8]. . 8
Figure 7: Longitudinal residual stress distribution in FSW 6013Al-T4 [1]. ....................................... 8
Figure 8: Demonstrates hardness profile in the weld region of a heat treatable alloy, AA6061-T6, as
a function of travel speed [7]............................................................................................................... 9
Figure 9: (a) Yield stress, (b) Ultimate stress (UTS), and (c) elongation of FSW samples of 7075Al
alloy across the weld [1].................................................................................................................... 10
Figure 10: S–N curves of base metal, FSW weld, laser weld and MIG weld for 6005Al-T5 [1]. .... 11
Figure 11: (a) A costum-built FSW machine [17], (b) an FSW robot [15] , and (c) a sturdy 5-axis
FSW machine [16]. ........................................................................................................................... 15
Figure 12: A schematic to illustrate: (a) front view of the FSW process with tilt angle, (b) side view
of the process [22]. ............................................................................................................................ 17
Figure 13: Schematic of: (a) Flared-Triflute tool, (b) A-skew tool developed by TWI from different
views [1]. ........................................................................................................................................... 18
Figure 14: Schematic of parameters of tool design [20]. .................................................................. 19
Figure 15: Effect of shoulder angle and diameter on axial force [20]............................................... 19
Figure 16: Schematic of resistance heating method [32]. ................................................................ 20
Figure 17: Axial force vs. tool rotational speed on 6.35 mm thick 6061-T6 and 2195-T6 friction stir
welded at 800 rpm, 120 mm/min, and a 1° tilt angle [26]. ................................................................ 21
Figure 18: Axial force vs. tool rotational speed on 6.35 mm thick 6061-T6 and 2195-T6 friction stir
welded at 800 rpm, 120 mm/min, and a 1° tilt angle [26]. ................................................................ 21
Figure 19: Effect of (a) tool rotational speed and (b) traverse speed on the axial forces during FSW
of AA 5083 [33]. ............................................................................................................................... 22
x
Figure 20: Influence of tool rotational and traverse speed on the axial force during FSW process on
3.07 mm thick AA6061-T6 at right angle [20]. ................................................................................ 22
Figure 21: Axial force vs. plunge depth on a 6.35 mm thick 6061-T6 and 2195-T6 friction stir welded
at 800 rpm, 120 mm/min, and a 1° tilt angle [26]. ............................................................................ 23
Figure 22: Influence of the shoulder angle and diameter on the UTS of AA6061-T6 sheets welded by
the RAFSW [20]. .............................................................................................................................. 24
Figure 23: Effect of (a) rotational speed and (b) traverse speed on UTS and hardness [35]. ........... 25
Figure 24: Effect of welding parameters on mechanical properties of AA1100 welded samples by
FSW [12]. .......................................................................................................................................... 26
Figure 25: Effect of FSW process parameters on mechanical properties of AA6061-T4 welded
samples by FSW [36]. ....................................................................................................................... 28
Figure 26: Relationship between axial force and (a) tool rotational speed, (b) traverse speed, and (c)
plunge depth with indication of the range of these parameters to obtain a defect-free weld [39]. ... 29
Figure 27: The domain of parameters to obtain a defect-free weld [39]. .......................................... 30
Figure 28: Schematic of the experiment [34]. ................................................................................... 30
Figure 29: The relation between the axial load and tensile strength [34]. ........................................ 31
Figure 30: Fatigue behavior of welded AA 6082-T6 samples at different traverse speeds [43]. ...... 32
Figure 31: Common mating variations in butt welding configuration; (a) gap, (b) mismatch, and (c)
misalignment [46]. ............................................................................................................................ 32
Figure 32: Joint efficiency vs. UTS for an industrial robot and a CNC machine using a 3° tilt angle
[22]. ................................................................................................................................................... 33
Figure 33: (a) Hardness profile at different traverse speed for as-welded samples, effect of water-
cooling (WC) and post weld artificial aging (AA) on samples welded at: (b) 100 rpm, (c) 400 rpm,
and (d) 800 rpm [48]. ........................................................................................................................ 34
Figure 34: Hardness profile of AA-6061-T6 welded samples at different rotational and traverse speeds
before and after the PWHT [49]........................................................................................................ 35
Figure 35. Tensile test results of (a) AA6061-T6 and (b) both alloys welded at different rotational and
traverse speeds before and after the PWHT [49]. ............................................................................. 35
Figure 36: Impact of PWHT on the hardness and UTS of the FSW joints welded at (a) different
rotational speed, (b) different traverse speeds [50]. .......................................................................... 36
Figure 37: Hardness profile of as-weld samples (AW), artificial aged (AA), and solubilized followed
by quenching and artificial aging (STA) [52]. .................................................................................. 37
Figure 38: Friction stir welded central tunnel assembly of the Ford GT which is made up of aluminum
stampings and extrusions [4]. ............................................................................................................ 38
xi
Figure 39: A structure consisting lap joint that is mainly assembled by riveting. ............................ 39
Figure 40: Lap joint configurations: (a) the upper sheet on the advancing side, (b) the lower sheet on
the advancing side [60]. .................................................................................................................... 39
Figure 41: Aluminum extruded profile to make railway cars body by FSW (1: before welding, 2: after
welding) [61, 62]. .............................................................................................................................. 40
Figure 42: Failure position and stress prediction around the joint [61]. ........................................... 40
Figure 43: An overview on categorized friction stir based technologies [64]................................... 41
Figure 44: (a) FSW tool and tool holder, (b) Welding set-up including the extruded bars, the clamping
system and the Kistler 3-axis dynamometer...................................................................................... 53
Figure 45: Schematic of tensile test samples..................................................................................... 53
Figure 46 : The architecture of the developed neural networks (NNs): (a) a 3-5-1 NN for UTS, (b) a
3-6-1 for downward axial force, (c) a 6-5-1 for UTS in presence of joint fit-up disturbances. ........ 59
Figure 47 : The effect of plunge depth on generated axial force and tensile properties of the welded
samples at optimized traverse speed of 540 mm/min, rotational speed of 3500 rpm, and various plunge
depths from 5.95 mm to 6.2 mm. ...................................................................................................... 60
Figure 48: Contour plots to show the correlation of the generated axial force during RAFSW process
with: (a) tool plunge depth and traverse speed (at constant rotational speed of 3500 rpm), (b) tool
plunge depth and rotational speed (at constant traverse speed of 540 mm/min), (c) tool rotational and
transverse speed (at constant tool plunge depth of 6.15mm). The contour plots extracted from final
ANN model for force. ....................................................................................................................... 61
Figure 49: Contour plots to show the correlation of the UTS of RAFSW joints with: (a) tool plunge
depth and traverse speed (at constant rotational speed of 3500 rpm), (b) tool plunge depth and
rotational speed (at constant traverse speed of 540 mm/min), (c) tool rotational and transverse speed
(at constant tool plunge depth of 6.15mm). The contour plots extracted from final ANN model for
UTS. .................................................................................................................................................. 62
Figure 50: Contour plots to show the correlation of the UTS of RAFSW samples with: (a) gap and
tool plunge depth, (b) mismatch and tool plunge depth, (c) misalignment and tool plunge depth, and
(d) thickness variations of the plates and tool plunge depth. All RAFSW samples are welded at
optimized tool rotational speed of 3500 rpm and traverse speed of 540 mm/min. The contour plots
extracted from final ANN model for UTS of welded samples in presence of joint fit-up disturbances.
........................................................................................................................................................... 64
Figure 51: Contour plots to show the correlation of the UTS of welded samples with two types of
disturbances at the same time which are: (a) mismatch and gap, (b) misalignment and gap, and (c)
misalignment and mismatch. All RAFSW samples are welded at optimized tool rotational speed of
xii
3500 rpm and traverse speed of 540 mm/min. The applied tool plunge depth was 6.2 mm. The contour
plots extracted from final ANN model for UTS of welded samples in presence of joint fit-up
disturbances. ...................................................................................................................................... 65
Figure 52: (a) Top surface view of the RAFSW joint welded at optimized condition consists of tool
plunge depth of 6.15 mm, rotational speed of 3500 rpm, and traverse speed of 540 mm/min, (b) macro
structure of the RAFSW joint which shows the nugget zone shape with unarmed eyes, (c) schematic
of FSW joint illustrates distinct microstructural regions called nugget zone (NZ),
Thermomechanically affected zone (TMAZ), heat affected zone (HAZ), and base material (BM). 66
Figure 53: Microstructure of (a) base metal, (b) HAZ, (c) TMAZ, and (d) NZ, at 500X magnification.
The RAFSW joint welded at optimized condition consists of tool plunge depth of 6.15 mm, rotational
speed of 3500 rpm, and traverse speed of 540 mm/min. ................................................................... 67
Figure 54: Hardness profile for the optimal sample along the top, center, and bottom lines as indicated
in Figure 52. Besides, the hardness profile of the PWHT sample along its centerline. The RAFSW
joint welded at optimized condition consists of tool plunge depth of 6.15 mm, rotational speed of
3500 rpm, and traverse speed of 540 mm/min. ................................................................................. 68
Figure 55: (a) The set-up of welding consist of the extruded bars, the clamping system and the Kistler
3-axis dynamometer; (b) FSW tool and tool holder; (c) top view of the welded sample by optimized
process parameters [70]..................................................................................................................... 74
Figure 56: Schematic of tensile test samples..................................................................................... 75
Figure 57: The effect of natural aging time on tensile strength: (a) and elongation; (b) of welded
samples at optimized process parameters by RAFSW. ..................................................................... 75
Figure 58: (a) Macrostructure at the cross-section of the joint marked with the distinct weld regions;
(b) Micro-Vickers hardness map around the welded region of a naturally aged sample made by
RAFSW technique operated at optimized process parameters.......................................................... 76
Figure 59: Microstructure of welded sample around the nugget zone at: (a) advancing side; (b) and
retreating side for naturally aged joints made by RAFSW at optimized process parameters. .......... 76
Figure 60: Microstructure of: (a) top; (b) middle; (c) and bottom of NZ zone; (d) base material; (e)
HAZ area far from weld region; (f) and HAZ area near NZ for naturally aged joints made by RAFSW.
........................................................................................................................................................... 77
Figure 61: The effect of artificial aging at: (a) 180°C; (b) 200°C; (c) 220°C; (d) and 240°C under
different aging times on the tensile strength of the RAFSW samples. .............................................. 78
Figure 62: Hardness distribution along the centerline of the cross-section of the joints for: (a) naturally
aged RAFSW sample; and for artificially aged samples at: (b) 160°C for 18h and; (c) 220°C for 30
min. ................................................................................................................................................... 80
xiii
Figure 63: Microstructure of welded sample around nugget zone at advancing side (a) and retreating
side (b) for RAFSW joint aged artificially at 160 °C for 16 h. ......................................................... 80
Figure 64: Microstructure of top (a), middle (b), and bottom (c) of NZ zone, base material (d), HAZ
area far from weld region (e), and HAZ area near NZ (f) for RAFSW joint aged artificially at 220 °C
for 30 min. ......................................................................................................................................... 80
Figure 65: The effect of solubilizing heat treatment followed by artificial aging at different
temperatures (at optimized durations obtained from the previous section) on the tensile properties of
the welded samples and plain material. ............................................................................................. 83
Figure 66: Microstructure around (a) and within (b) the nugget zone of the welded sample after W+T6
heat treatment. ................................................................................................................................... 84
Figure 67: (a) The tool shape to make lap joints by RAFSW technique, (b) the close view of the tool
and tool holder, (c) the RAFSW set-up including the tool, tool holder, clamped sheets on the back-
plate, and the back-plate installed on the dynamometer, (d) the welded sample, (e) the schematic of
the tool and the design parameters (the pin lead, not illustrated, is the distance between the threads on
the pin). ............................................................................................................................................. 91
Figure 68: The schematic of the welding process for two possible configurations of the lap joints. 92
Figure 69: the schematic and dimensions of the weld coupons to make the tensile test. (a)
Configuration No. 1, when the advancing side is on the upper sheet (b) configuration No. 2, when the
retreating side is on the upper sheet. ................................................................................................. 92
Figure 70: (a) 16 tools made according to L16 DOE presented in Table 21, (b) 4 tools made regarding
Table 22............................................................................................................................................. 93
Figure 71: The effect of learning rate and momentum coefficient on (a) root-mean squared error
(RMSE) and (b) maximum error when the architecture of the model is 10-8-1. .............................. 96
Figure 72: The architecture of ANN models in this paper. The architecture of the model for downward
axial force and fracture force is 10-8-1 for both of them. ................................................................. 96
Figure 73: Contour plots of: (a) downward axial force during RAFSW process, and (b) failure force
at tensile shear test versus shoulder diameter, SD, and shoulder groove depth, SGD, while other
geometry and process parameters are the same as parameters of sample No. 21. The contour plots are
extracted from developed ANN models in this paper. ...................................................................... 99
at tensile shear test versus pin length, PL, and pin angle, PA, while other geometry and process
parameters are the same as parameters of sample No. 21 except for the plunge depth wich is the same
as pin length. The contour plots are extracted from developed ANN models in this paper. ........... 100
xiv
at tensile shear test versus pin base diameter, PBD, and pin lead, PLD, while other geometry and
process parameters are the same as parameters of sample No. 21. The contour plots are extracted from
developed ANN models in this paper. ............................................................................................ 101
at tensile shear test versus welding traverse speed, V, and welding rotational speed, w, while other
geometry and process parameters are the same as parameters of sample No. 21. The contour plots are
extracted from developed ANN models in this paper. .................................................................... 102
at tensile shear test versus tool plunge depth, PD, and welding traverse speed, V, while other geometry
and process parameters are the same as parameters of sample No. 21 except for the plunge depth
which is the same as pin length. The contour plots are extracted from developed ANN models in this
paper. ............................................................................................................................................... 103
at tensile shear test versus tool plunge depth, PD, and welding rotational speed, w, while other
geometry and process parameters are the same as parameters of sample No. 21 except for the plunge
depth wich is the same as pin length. The contour plots are extracted from developed ANN models
in this paper. .................................................................................................................................... 103
Figure 79: Illustration of the application: (a) truck box, (b) side panels, (c) dimension of the delta-
shaped stiffener. .............................................................................................................................. 108
Figure 80: (a) The tool to make lap joints by RAFSW, (b) The close view of the tool and tool holder,
(c) The RAFSW set-up including the tool, tool holder, clamped sheets on the back-plate and the back-
plate installed on the dynamometer, (d) The welded sample [1]. ................................................... 110
Figure 81: The schematic and dimensions of the weld coupons to make the tensile test. .............. 110
Figure 82: The architecture of developed ANN models for the downward axial force and fracture
force (4-9-1). .................................................................................................................................. 113
Figure 83: Contour plots of the DAF versus: (a) traverse speed and rotational speed (when the plunge
depth is 2.65 mm), (b) plunge depth and rotational speed (when the traverse speed is 1700 mm/min),
(c) plunge depth and traverse speed (when the rotational speed is 4400 rpm). The contour plots are
extracted from developed ANN models in this study. .................................................................... 115
Figure 84: Contour plots of the FF versus: (a) traverse speed and rotational speed (when the plunge
depth is 2.65 mm), (b) plunge depth and rotational speed (when the traverse speed is 1700 mm/min),
(c) plunge depth and traverse speed (when the rotational speed is 4400 rpm). The contour plots are
extracted from developed ANN models in this study. .................................................................... 115
xv
Figure 85: (a) The new tool with the tapered entry, (b) the spring-loaded tool holder, (c) the whole
set-up for the RAFSW in force control on Fryer MC15 CNC machine, and (d) a welded sample at the
optimized welding condition. .......................................................................................................... 117
Figure 86: Welding process of a delta-shaped structure clamped partially on one side on the MC15
machine. The part is installed on a dynamometer plate: (a) The clamped structure on one side. There
is a considerable gap between the sheets in the other side, (b) One side is welded. The other side is
under welding (the gap is compensated; and the sheets are squeezed together), (c) The welded sample
in both sides..................................................................................................................................... 118
Figure 87: (a) The RAFSW process in force control mode to weld a large truck side panel on a big-
size CNC router, (b) the welded samples at 3 different welding sequences with and without weld tags,
(c) the top view of the clamped and tagged panels (the location of tags and clamps are shown in
yellow and orange, respectively), (d) distortion distribution of each sample (e) schematic of the
procedure with the identification of the advanced side (AS) and retreating side (RS) of each weld
path, (f) the macrostructure of welded samples. ............................................................................. 119
Figure 88: The stress distribution (Von-misses criterion) for a 3-inch-long weld coupon under a
4448.22 N tensile load..................................................................................................................... 120
Figure 89: (a) the experimental set-up to test weld coupon under static load in tension, (b) the broken
sample after testing, (c) a close look to the broken area and the crack initiation site in the weld area.
......................................................................................................................................................... 121
Figure 90: The stress distribution for a 1-ft-long weld coupon under a (a) 1200 N and (b) 2880 N
compressive load. The magnified, cross-sectioned view of the joint is shown both in the corner and
middle of the assembled parts. The amount of tensile stresses in the the crack-like area of the joint
are identified in these cross-sections. .............................................................................................. 122
Figure 91: the experimental set-up to test weld coupon under dynamic load in compression. ....... 123
Figure 92: (a) the broken sample after fatigue test, (b) as can be seen, maximum plastic deformation
is located in the middle of the weld seam; the crack initiated from the middle of weld area, (c) a close
look to the broken corner demonstrates that the crack initiated from the crack-like region of the weld.
......................................................................................................................................................... 123
Figure 93: The schematic of the multilayer Perceptron .................................................................. 136
Figure 94: The architecture of ANN models in this paper. The architecture of the model for both
downward axial force and fracture force is 10-8-1. ........................................................................ 140
xvi
List of Tables
Table 1: Significant advantages of FSW over fusion welding [1]. ..................................................... 1
Table 2: A list of the FSW tool design and process parameters for some aluminum alloys [14]. .... 12
Table 3: Some FSW tool designs [2]. ............................................................................................... 13
Table 4: Optimized FSW process parameters for different alloys [12]............................................. 25
Table 5: The percentage of UTS improvement by PWHT [50]. ....................................................... 36
Table 6: The first DOE. ..................................................................................................................... 55
Table 7: The second DOE ................................................................................................................. 56
Table 8: The third DOE (P.D.=6.15 mm) ......................................................................................... 56
Table 9: 4th dataset of experiments to enrich the ANN model. ........................................................ 56
Table 10: Some experiments to explore about the effect of plunge depth on force and UTS for joints
welded at optimal tool rotational and traverse speeds. ...................................................................... 56
Table 11: Confirming tests for the ANN model. ............................................................................... 56
Table 12: The DOE for the effect of joint fit-up disturbances. ......................................................... 57
Table 13: Confirming tests for the ANN model including part......................................................... 57
Table 14: Training parameters used for developed ANN models. .................................................... 58
Table 15: Error values for the final ANN models. ............................................................................ 59
Table 16: Correlation of the final ANN models. ............................................................................... 59
Table 17: Efficient working window of the process parameters of the RAFSW technique ............. 63
Table 18: Tensile properties and joint efficiency of the welds aged naturally and at optimized artificial
aging conditions. ............................................................................................................................... 78
Table 19: Tensile strength of the welded samples subjected to natural aging for less than one day to
21 days prior to artificial aging process conducted at different conditions based on the obtained results
in the previous part of the study. ....................................................................................................... 82
Table 20: Chemical composition and tensile strength of base metal. ............................................... 91
Table 21: The experiments conducted according to L16 orthogonal array. For each condition of the
L16 array, tow tests are done with different tool plunge depths. ...................................................... 93
Table 22: Some more experiments conducted to explore more regarding the effect of tool and process
parameters on downward axial force and fracture force. .................................................................. 95
Table 23: The details regarding the developed ANN models in this paper to study the effect of tool
geometry and process parameters on downward axial force and failure force. ................................ 97
Table 24: The RMSE and maximum error for the training data, the confirmation data, and the overall
data when the architecture of the neural networks is 10-8-1. ............................................................ 98
xvii
Table 25: The amount of different kind of errors for the developed ANN models........................... 98
Table 26: The amount of correlation coefficient (R2) for the training data, confirmation data, and
overall data for the developed ANN models. .................................................................................... 98
Table 27: The amount of experimental data, predicted data, and its error for the confirmation
experiments. ...................................................................................................................................... 98
Table 28: An efficient working window of the geometry and process parameters to make lap joints
by RAFSW. ..................................................................................................................................... 103
Table 29: Chemical composition and tensile strength of the sheet metal. ...................................... 109
Table 30: The characteristics of the optimized tool used in this research. ...................................... 110
Table 31: The experiments conducted according to the L9 orthogonal array. ................................ 111
Table 32: The additional tests. ........................................................................................................ 111
Table 33: The confirming tests........................................................................................................ 111
Table 34: The optimized experiment. ............................................................................................. 111
Table 35: The details of the developed ANN models for both the downward axial force and failure
force................................................................................................................................................. 112
Table 36: The RMSE and maximum error for the training, confirmation, and entire set of data ... 113
Table 37: The amount of different kinds of errors for the final developed ANN models. .............. 113
Table 38: The amount of correlation coefficient (R2) for the training, confirmation and entire set of
data. ................................................................................................................................................. 113
Table 39: The amount of experimental and predicted data and their errors for the confirmation and
optimized experiments. ................................................................................................................... 114
Table 40: An efficient working window of the process parameters. ............................................... 115
Table 41: The calculated errors for the validation data predicted by some single hidden layer ANNs
with different number of neurons in the hidden layer. .................................................................... 140
Table 42: The details of the developed ANN models. .................................................................... 141
Table 43: the input and output for trained and validation data for fracture force using a 10-8-1 ANN.
......................................................................................................................................................... 141
Table 44: calculated errors and correlation coefficients for trained, validation, and entire set of data
for fracture forces modeled by 10-8-1 ANN model. ....................................................................... 142
xviii
Dedication
To my beloved husband and son
xix
Acknowledgment
I acknowledge my advisor and mentor, Prof. Michel Guillot. His knowledge, drive, dedication to
student development, guidance, and commitment to excellence are both inspiring and remarkable. He
provided me with the opportunity to sharpen my skills in the field of product development and
manufacturing processes, specifically through the related courses provided by him and innumerable
discussions that we had with regard to my Ph.D. project. Moreover, being his student and working in
his labs as a member of PI2/REGAL research team allowed me to be involved in many invaluable
discussions regarding the academic and industrial projects of the team. That atmosphere gave me the
chance to ponder about many interesting challenges and sharing ideas from not only an industrial lens
but also a meticulous scientific perspective. More importantly, I have learned to think beyond the box
when facing challenges. For all of these, I am eternally grateful to Prof. Michel Guillot. I also thank
him for his financial support during my Ph.D. program.
I also thank the other members of my thesis committee, Prof. Alain Curodeau, Prof. Augustin
Gakwaya, and Prof. Abderrazak El Ouafi for their encouragement, guidance and insightful
commentary on my work. I am deeply grateful and appreciative of the time and energy that they have
allocated to my thesis. I also thank Prof. Benoît Lévesque, the director of the graduate students of the
mechanical engineering department, for his encouragement and support. Furthermore, I thank the
director and technicians of the department’s machine shop for their help and support throughout my
project. Additionally, I would like to thank the members of PI2/REGAL research team who have made
this research possible.
I must thank my parents for their unconditional love, constant encouragement, and support. I am
grateful of the thoughtful tips that my father gave me in the milestones of my life particularly with
respect to my education. I also wish to especially thank my husband, Tohid, for his great love,
understanding, and support throughout my education. He is definitely my biggest cheerleader and has
always inspired me by his faith in my abilities. Finally, I thank my son, Aiden William, who has just
been added to our family. His existence has doubled my perseverance for seeking success in my
career to be inspiring for him in the future.
xx
Preface
This doctoral thesis is presented to the department of mechanical engineering at Laval University.
This research project was carried out under the supervision of Prof. Michel Guillot. In this thesis, a
friction stir welding technique operated at right angle is developed to make the process low-cost and
industrially desirable. This thesis consists of nine chapters. The first chapter introduces the friction
stir welding (FSW) process. The second chapter provides the literature review with regard to the
studied aspects in the next chapters of this thesis. The third chapter states the problem, objectives,
and the general methodology of the thesis. The following four chapters, chapters 4 to 7, present the
major achievements of the current study which lead to four scientific papers. The last chapter presents
the conclusion and perspectives for future works. In addition to the mentioned papers as major
contribution of the student, there are two other papers, paper 5 and 6, as minor contribution.
The papers and the authors’ contributions are as followed:
Paper 1 (Chapter 4): Development of friction stir welding technique at right angle (RAFSW) applied
on butt joint of AA6061-T6 aluminum alloy
Authors: Mahboubeh Momeni, Michel Guillot
Journal: The International Journal of Advanced Manufacturing Technology, Vol 99, 3077–3089
(2018).
Author Contributions: M.M. and M.G. conceived and designed the experimental processes and made
the welds; M.M. performed the characterization tests, M.M. conducted the ANN modeling and the
analysis of the results; M.M. wrote the manuscript; M.G. supervised the experiments, modeling, and
analysis of the results, and revised the manuscript.
Paper 2 (Chapter 5): Post-Weld Heat Treatment Effects on Mechanical Properties and
Microstructure of AA6061-T6 Butt Joints Made by Friction Stir Welding at Right Angle (RAFSW)
Journal: Journal of Manufacturing and Materials Processing (Open access), Vol 3 (2), 42 (2019).
Author Contributions: M.M. and M.G. made the welds; M.M. and M.G. conceived and designed the
PWHT processes and the characterization; M.M. performed the experiments, the characterization
tests, and the analysis of results; M.M. wrote the manuscript; M.G. supervised the experiments and
the analysis; M.G. revised the manuscript.
xxi
Paper 3 (Chapter 6): Effect of Tool Design and Process Parameters on Lap Joints Made by Right
Angle Friction Stir Welding (RAFSW)
Journal: Journal of Manufacturing and Materials Processing (Open access), Vol 3 (3), 66 (2019).
the welds; M.M. performed the characterization tests, M.M. conducted the ANN modeling and the
analysis of the results; M.M. wrote the manuscript; M.G. supervised the experiments, modeling, and
analysis of the results, and revised the manuscript.
Paper 4 (Chapter 7): Implementation of Right Angle Friction Stir Welding (RAFSW) to Assemble
the Side Panels of Truck Box
Journal: The International Journal of Advanced Manufacturing Technology, accepted to publish,
July 2020
the tools and welds; M.M. performed the characterization tests, M.M. conducted the ANN modeling
and the analysis of the results; M.M. conducted the finite element modeling, M.M. wrote the
manuscript; M.G. supervised the experiments, modeling, and analysis of the results, and revised the
manuscript.
Paper 5 (minor contribution): Developing a new manufacturing technique, which is an integration

of machining and RAFSW technique operated on a CNC milling machine
Authors: Michel Guillot, Mahboubeh Momeni
Manuscript is under writing to send to a peer-reviewed journal.
Author Contributions: M.G. and M.M. conceived and designed the experimental processes; M.G. and
M.M. have done some preliminary tests to use cutting fluid for FSW; M.G. prepared the CNC
program and made the samples; M.G. and M.M. discussed the results; M.G. writes the manuscript.
Paper 6 (minor contribution): Comparing the mechanical properties of optimized robotized pulsed
MIG welded butt joints with the joints made by friction stir welding at right angle
Authors: Mahboubeh Momeni, Milad Bahrami, Michel Guillot
Manuscript is under writing to send to a peer-reviewed journal.
Author Contributions: M.G, M.B, and M.M. conceive and design the experimental processes and
made the welds; M.G, M.B, and M.M. perform the characterization tests and the analysis of the
results; M.M. writes the manuscript; M.G. revises the manuscript.
xxii
Introduction
General background
In recent decades, friction stir welding (FSW) is remarked as one of the most considerable
developments among metal joining processes. FSW was invented by Wayne Thomas at The Welding
Institute (TWI) of UK in 1991 where it is used to join aluminum alloys at first. The FSW process
provides users with many advantages such as higher energy efficiency, versatility and being
environmentally friendly compared to fusion welding techniques. It can be used to join non-weldable
aluminum alloys, dissimilar alloys, and composites which their joint properties are not acceptable by
fusion welding techniques. It can be employed to join a wide range of low melting point alloys such
as aluminum, magnesium, and copper alloys, and also to weld some high melting point alloys such
as steels. [1-3].
Benefits and challenges

The FSW process is considered a green technology, as it is an environmentally friendly and energy-
efficient process. Since FSW is a solid-state process, it consumes less energy compared to
conventional fusion welding techniques. In addition, this welding technique does not need to use
shielding gas, powder or flux. Furthermore, there is not any chemical composition change in the
welded region of parent metal due to using a non-consumable rotating tool instead of consumable
electrodes [1, 2, 4]. Shortly, the most crucial benefits of FSW over fusion welding are summarized in
Table 1 [1].
Table 1: Significant advantages of FSW over fusion welding [1].
Metallurgical benefits Environmental benefits Energy benefits

Solid phase process No shielding gas required Improved materials use (e.g. jointing
Low distortion of workpiece No surface cleaning required different thicknesses) allows reduction in
Good dimensional stability Eliminate grinding wastes weight
Good repeatability Eliminating solvents for degreasing low energy consumption compared to
No loss of alloying elements Consumable materials other welding techniques specially laser
Excellent metallurgical properties welding
Fine microstructure Decreased fuel consumption in light
Absence of cracking weight transportation means like trains
Although the FSW process is almost a new welding technique, it is going to find its niche at an
accelerated pace in various industries such as aerospace, automotive, maritime industries, and
infrastructures such as bridges and railway structures [2, 4]. Despite the prominent advantages of the
FSW over other fusion welding techniques, this method is not explored completely in detail, yet.
1
Nowadays, there is a growing interest among researchers to develop and explore different aspects of
this new green metal joining technique to make it feasible and reliable for widespread use [1, 2, 4].
In this regard, many challenges have remained to be solved. For instance, it is needed to develop FSW
techniques to make them cost-effective for industrial use. Moreover, it is of industrial importance to
find efficient welding parameters for different materials, geometries, and configurations, and to
improve the tool design, fixture design, and controlling techniques to improve the welding efficiency.
[1, 2, 4]. Finding feasible solutions to the current challenges would pave the way to implementing
this welding method instead of fusion welding techniques in various industries, widely.
2
Literature Review
1.1 FSW Process and the joint specifications
The FSW is a solid-state metal joining process. During the FSW process, the material undergoes
intense plastic deformation at elevated temperatures using a non-consumable rotating tool. The tool
consists of a designed shoulder and a pin made of a harder material than the parent metal. The pin of
the rotating tool plunges into the abutting edges of the plates of the parent metal and traverse along
the joint’s line. The combination of these two types of movements, rotational and transverse move,
leads to transporting the material from the front of the pin to the back of that. Also, it causes heating
of the material via plastic deformation and friction between the tool and the workpiece. Half of the
welded plate is called advancing side when the direction of rotation and welding is the same. The
other half would be retreating side [1, 2, 4]. Figure 1.a depicts a schematic of the FSW process in a
butt-joint configuration [2]. Figure 1.b and c depict an FSW joint and FSW tool which consists of a
pin and shoulder are, respectively [2].
Figure 1: (a) Schematic of FSW in a butt-joint configuration, (b) FSW seam weld, (c) an FSW tool, which
consists of a pin and shoulder [2].
3
1.1.1 Joint Configuration
The FSW is performable in almost all joint configurations [1]. Figure 2 shows the different FSW joint
configurations including square butt, edge butt, T butt, lap, and fillet joints. The samples do not need
any special preparation process, such as beveling or cleaning by chemicals, to weld by the FSW. This
feature is desirable for industrial applications [1].
Figure 2: FSW joint configurations: (a) square butt, (b) edge butt, (c) T butt, (d) lap butt, (e) multiple lap, (f)
T lap, and (g) fillet joint [1].
1.1.2 Heat input

The generated heat input during the FSW process is estimated and modeled by different methods [1].
Formula 1.1 shows the relationship between the heat input, welding parameters, the tool geometry,
and the materials properties of the base metal:
2𝜋
𝑞 = ( 3𝑆 ) 𝜇𝐹𝑁𝑅𝜂, (1.1)
where q is the heat input (kJ/mm); S is the welding traverse speed (mm/min); F is the downward axial
force (kN); N is the tool rotational speed (rpm); R is the tool pin radius; 𝜇 is the coefficient of friction;
and 𝜂 is the welding efficiency.
As seen in equation 1, the heat input decreases by increasing the welding traverse speed. Figure 3
demonstrates this fact, experimentally [5]. It is clarified that the quality of the weld could reduce and
the distortion increases by increasing the heat input [6].
4
Figure 3: Heat input vs. welding speed input [6].
1.1.3 Microstructural Evolution

Aluminum alloys are classified into two categories called heat treatable and non-heat-treatable alloys.
In this connection, 1xxx, 3xxx, and 5xxx alloys are categorized as non-heat-treatable alloys. In other
words, they are called non-precipitation-hardening alloys which means they do not strengthen by
second-phase particles. On the other hand, heat treatable alloys, which called precipitation-hardening
alloys, are classified as 2xxx, 4xxx, 6xxx, and 7xxx aluminum alloys [5]. The microstructure of these
alloys undergo some changes by the FSW process. The FSW process which includes intense plastic
deformation at elevated temperatures causes the formation of three microstructural zones across the
weld region. They are called nugget (stirred) zone (NZ), thermo-mechanically affected zone (TMAZ),
and heat-affected zone (HAZ) [1, 5]. Figure 4 depicts these microstructurally different regions in
friction stir welded aluminum alloys [7].
The NZ is a recrystallized fine-grained region since it is under intense plastic deformation at elevated
temperatures during the FSW process. Since the shape of this region depends on the process
parameters, it is classified into two types called basin-shaped and oval-shape regions. From a
microstructural perspective, there are three important aspects to be considered in the NZ that are grain
size, precipitates’ size, and their distribution. They play a prominent role in the mechanical properties
of the welded joints by FSW technique [1, 2, 5, 7]. Because of dynamic recrystallization in the NZ,
grains are fine and equiaxed, typically. All the process parameters besides the cooling rate are of
paramount importance in determining the grain size of the NZ [1, 5]. Coarsening or dissolving of
precipitates happens in this region as the temperature reaches 400 to 550°C in the NZ during FSW of
aluminum alloys [1].
5
Figure 4: (a) Typical microstructure of friction stir welded aluminum alloys, (b) higher magnification in the
retreating side, and (c) higher magnification in the advancing side [7].
The linking region between the NZ and the HAZ is the TMAZ. This unique feature is what
differentiate FSW from fusion welding. The material in the TMAZ is highly deformed and exposed
to high temperatures. However, it does not undergo recrystallization, as the applied strain is not
adequate. It can be seen that some precipitates are dissolved or coarsened in the TMAZ because of
exposing to high temperatures [1, 5].
The last affected zone is HAZ, which undergoes just a thermal cycle. Generally, for the heat treatable
aluminum alloys, HAZ is a region that experiences high temperatures. The temperature within this
region is high enough to affect precipitate size and distribution, notably [1, 5]. Figure 5 depicts the
size, distribution, and shape of precipitates inside the grains and along grain boundaries in the parent
metal, HAZ and TMAZ for a 7075Al-T651 which is welded by FSW at the tool rotational speed of
350 rpm and traverse speed of 15 mm/min. It is obvious that the FSW process considerably affects
the precipitates’ size, distribution, and shape [1].
6
Figure 5: Precipitate size, distribution, and shape in: (a) base metal, (b) HAZ, (c) TMAZ near HAZ, and (d)
TMAZ near nugget zone [1].
1.1.4 FSW Joints Defects

During FSW, some defects may occur across the weld zone because of the flowing pattern of metal
during the process. Process parameters, preheating, and cooling rate can affect the generation of these
defects. Therefore, a defect-free friction stir welded joint can be achieved by controlling the process,
carefully [1, 8]. It is demonstrated that the tool traverse speed increase causes the generation of
wormhole defects. In another research, it is reported that the ratio of tool rotational speed to tool travel
speed plays a key role in obtaining a defect-free welded joint [2]. Figure 6 demonstrates the probable
defects in the FSW joints. From the presented curve in this figure, it is concluded that there is a
suitable range of the rotational speed and traverse speed to obtain a defect-free weld. It is of great
practical importance to explore this range for various alloys, thicknesses, and welding configurations
[8].
7
Figure 6: FSW common defects and the safe range of tool rotational speed and traverse speed [8].
1.1.5 Residual Stresses

Evaluation of the developed residual stresses in the friction stir welded samples is necessary to control
the post-weld mechanical properties of the welded part, especially its fatigue behavior [9, 10].
According to Figure 7, longitudinal residual stress in the weld zone has an M-shaped distribution
across the weld. It elucidates that the HAZ sustains the maximum tensile longitudinal stresses [1].
Figure 7: Longitudinal residual stress distribution in FSW 6013Al-T4 [1].
8
1.1.6 Hardness
In heat-treatable aluminum alloys, the FSW process causes softening of the NZ that is mainly due to
the coarsening and dissolution of precipitates rather than the change of grain size. It is reported that
the size, distribution, morphology, and shape of precipitates besides dislocation density changes by
the FSW process, significantly. As a result, these considerable changes cause prominent changes in
the materials hardness and the mechanical properties around the joint [1, 11]. Regarding this issue,
Figure 8 demonstrates a W-shaped hardness profile in the weld region of a heat treatable alloy, 6061-
T6 Al, at different traverse speed [7].
For FSW joints of non-heat-treatable alloys, Hall-Petch relationship could not define their hardness
profile, while Orowan strengthening theory could explain it. It suggests that the hardness profile is
related to the relative strengthening contribution of grain boundaries, substructure, and strengthening
particles [1].
Figure 8: Demonstrates hardness profile in the weld region of a heat treatable alloy, AA6061-T6, as a function
of travel speed [7].
1.1.7 Tensile Properties

A prominent change happens to the tensile properties and the fatigue behavior of a friction stir welded
alloy that originated from substantial microstructural evolution and developed residual stresses within
and around the FSW joints [1, 7]. Figure 9 demonstrates the tensile properties across the FSW joint
in a heat-treatable aluminum alloy [1]. It is elucidated that the yield stress is almost related to the
hardness. In addition, the fracture location and the location of the lowest hardness in the W-shaped
hardness profile seem are linked together [7].
9
(a)
(b)
(c)
Figure 9: (a) Yield stress, (b) Ultimate stress (UTS), and (c) elongation of FSW samples of 7075Al alloy
across the weld [1].
1.1.8 Fatigue behavior

The fatigue behavior of the welded structures and components is of paramount importance in the vast
majority of engineering applications. The FSW structures are prone to initiate fatigue cracks and
fracture from the welded joints and its surroundings. Hence, identifying the fatigue behavior of the
friction stir welded alloys, in terms of S-N curves, is of great practical importance [1, 7]. Figure 10
illustrates the fatigue behavior of a welded aluminum alloy using different welding techniques. As
can be seen, welding has a detrimental effect on the fatigue behavior of the parent metal, but the FSW
causes a less negative impact on the fatigue properties of the welded part compared to fusion welding
techniques like MIG welding [1].
10
Figure 10: S–N curves of base metal, FSW weld, laser weld and MIG weld for 6005Al-T5 [1].
1.2 Welding parameters

One of the main challenges that every manufacturer faces when it wants to utilize a welding method
is selecting the proper welding parameters and tools so that the welded joint meets the desired
specifications and properties [12]. In this regard, the FSW parameters including the process
parameters and the tool design have a great impact on the generated axial forces, mechanical and
physical properties of the welded joints. The most important process parameters are the tool rotation
speed and its direction, clockwise or counter-clockwise, traverse speed, tilt angle, which is the tool
angle regarding the workpiece surface, and plunge depth of the tool pin into the workpiece [1, 2]. The
generated heat during the FSW via intense plastic deformation and friction results from the rotational
and traverse movements of the tool into the workpiece. [1, 13].
In summary, process parameters including welding parameters and tool design parameters have a
great impact on the mechanical and physical properties of welded joints. Hence, they should be
selected properly to accomplish a sound, defect-free weld. Table 2 shows a list of the FSW tool design
and process parameters useful for some aluminum alloys, thicknesses, and configurations [14].
1.3 Tool Design

The FSW tool is made up of a shoulder and a pin. The tool heats material locally and controls the
flow of the material in the workpiece during the welding process [1]. Tool design significantly affects
the heat generation, flow behavior of material, quality, and uniformity of the welded joint [2]. During
the FSW process, the rotating tool plunges into the workpiece and traverses along it. Due to these two
kinds of motions, the heat will be generated via intense plastic deformation and friction in the
11
workpiece. If the tool plunges more into the material and the shoulder contacts the workpiece, more
heat generates and the tool prevents material from escaping the workpiece surface [1, 13]. In this
regard, it is shown that the vast majority of the heat input of the FSW process of thin plates is due to
deformational and frictional heating by the shoulder. However, it is mostly because of pin movements
in thick plates [13].
In general, there are three types of shoulders including concave, convex, and flat shoulders. To
increase the efficiency of these shoulders, they are developed by adding some features such as scrolls.
There is a lot of research focused on the developing features of the shoulders. In addition, many
efforts are made on improving the pin design by adding different features such as flutes [1, 2, 13].
Table 3 categorizes some designed FSW tools [2].
Table 2: A list of the FSW tool design and process parameters for some aluminum alloys [14].
12
Table 3: Some FSW tool designs [2].
1.4 FSW machines

Industrially, we can find three different categories of the FSW machines: (1) custom-design machines
for specific welding geometries, usually for serial production, (2) high capacity robots that
instrumentation and control are adapted to the FSW process, and (3) sturdy 4 and 5-axis CNC
machines designed and controlled in position and force especially made for the FSW process. Figure
11 illustrates examples of these machines [15-17]. Furthermore, researchers with limited budgets
sometimes use CNC and conventional machining machines for R&D purposes not for industrial
production. They employ machining machines only to operate in position control. The characteristics
of a machine and its capabilities are key aspects to determine which one is a suitable choice for a
certain FSW process. Some of these capabilities are high axis force, high stiffness, good accuracy,
and sensing ability [15, 18]. The mentioned FSW machines provide users with a wide range of
technical capabilities for different applications. They cost roughly from 500,000$ to multimillion
dollars.
Generally, FSW machines have the high force and stiffness capabilities. It is difficult to justify them
from an economic standpoint, particularly when they are considered being used only as an alternative
for a welding process. This type of machine could be a useful alternative solution when there are no
other means to assemble a component or the existing process for assembly is very expensive [18].
13
The FSW robots are not made specifically for the FSW process. In fact, they are adapted for the FSW
process. The drawback of the FSW robots is their limitation to weld thick plates. The current FSW
process in the industry needs high levels of force and moderate to a high level of stiffness. Hence,
FSW robots could be economic alternatives for costume-built machines to weld the plates with a
thickness of less than 6 mm.
Another type of machines is sturdy 4 and 5-axis CNC machines designed and controlled in position
and force, especially for the FSW process. They can sustain very high level of forces and can work
at tilt angle. Moreover, they can weld thick joints. However, these machines are very expensive.
As it is mentioned before, it is possible to use the machining centers as FSW machines after modifying
them. Because the FSW process instinctively has similarities with the machining process. Nowadays
developed FSW methods to use in the industry needs prominently higher forces than machining
processes. The vast majority of the FSW processes are done by a tilted tool to accomplish a sound,
defect-free weld. To deal with this situation, it is necessary to utilize 5-axis machining centers which
are costly or implementing some mechanical fixed solutions to 3-axis machines to overcome this
limitation. However, modifying these machines is relatively costly; and the machines for this purpose
should have higher capabilities than a conventional CNC machine for machining [18]. Thus,
improving the economic aspects of the FSW processes would make this joining process a great
alternative to conventional fusion welding methods.
14
(a)
(b)
(c)
Figure 11: (a) A costum-built FSW machine [17], (b) an FSW robot [15] , and (c) a sturdy 5-axis FSW
machine [16].
15
1.5 Making the FSW process affordable
As mentioned before, the affordability of a technology plays a crucial role in the widespread
implementation of it, especially when there are long-term established methods instead of the newly
introduced technique. Taking this into account, improving cost-effective FSW techniques is
significantly important to make the process affordable for industrial users [8, 19]. Generally,
manufacturers consider economic issues, cost, and productivity to select a production method. There
are some technical factors to select a proper FSW machine which includes force, stiffness,
intelligence, flexibility, and sensing requirements. These factors affect the cost of the machines and
the production process [18]. Developing reliable FSW techniques that need less equipment
capabilities could pave the way for widespread industrial use of the FSW process instead of fusion
welding techniques such as MIG welding. To this end, it is necessary to study the effect of tool design
and process parameters on the generated axial forces and the quality of the joints by a certain type of
machine. Afterwards, one can develop a reliable, cost-effective FSW technique.
Developing FSW techniques that could reduce the generated axial forces during the process could
decrease the needed load capacity and stiffness requirement of forecasted FSW machines. Another
aspect to reduce the required capabilities of the FSW machines is related to the tilt angle. Welding
using a tilted tool requires 5-axis or custom machines, while welding using a zero travel angle could
be done by a simpler 3-axis machine tool [20]. Therefore, the needed capabilities for the FSW
machine could be reduced, significantly, by developing FSW techniques operated at a right angle.
Hence, reducing the needed force and stiffness capabilities of the machines, besides utilizing 3-axis
machine tools without necessary modifications for tilt angle applications would result in a
considerable decrease in the cost of machines. Consequently, it provides the possibility of taking
advantage of conducting the FSW process by cheap CNC-machines which are widely present in the
current industries.
1.6 Study the effect of tool design and process parameters

1.6.1 Effect of tool tilt angle on generated axial force
The tool tilt angle plays a critical role in the quality of the FSW joints [21]. Figure 12 portrays the
schematic of the front view and side view of a FSW process operated at tilt angle [22]. Conducted
research up to now mainly demonstrates that a non-zero tool tilt angle should be applied to accomplish
a sound defect-free weld [18]. 4 or 5-axis or modified 3-axis machines are needed to apply a certain
tilt angle. This causes the increase of the machine’s cost, greatly [18].
16
Welding using a tilt angle limits the maneuverability of the FSW process. For example, it is essential
to utilize a zero tilt angle to weld complex non-linear patterns by FSW [23, 24]. Considering these
issues into account, developments regarding utilizing zero working angle can make a great change in
the needed capabilities of the FSW machines. Consequently, developing and optimizing the FSW
process at zero tilt angle, called right angle, can open a door to make the FSW process more feasible
for industry [8, 18, 19].
Figure 12: A schematic to illustrate: (a) front view of the FSW process with tilt angle, (b) side view of the
process [22].
It is demonstrated that the higher travel angle, the more axial force is generated to achieve a sound
weld [25]. Regarding this issue, Mehta and Badheka have shown that to weld a AA6061-T651 plate
to a Copper plate the axial force increases by the increment of the tilt angle, significantly [25].
Another research was done regarding the effect of the tilt angle on the quality of the welded joints of
6.5 mm thick AA 7075-T651. It was concluded that a 2° tilt angle must be applied to accomplish a
sound, defect-free weld. It was not possible to obtain quality welds with 0 and 1° tilt angles [21].
Melendez et al. demonstrated that by a 1° tilt angle, the amount of axial force to weld 6.35 mm thick
Al 6061-T6 is 1.3 kN higher than the force for welding with a zero tilt angle. It is mainly due to the
fact that by a 1° inclined angle there would be an additional pushing up longitudinal force. It should
be compensated by an additional amount of downward axial force to keep a certain plunge depth [26].
It is demonstrated that the higher travel angle, the more axial force is generated during the FSW
process to achieve a sound weld [25, 26]. Therefore, FSW machines with higher load capacity and
stiffness would be needed to weld with a tilt angle compared with welding by zero travel angle.
Additionally, the more the thickness of the plate, the more the axial force generates during the
welding. Thus, a certain FSW machine with certain characteristics can weld thicker plates by zero
travel angle than tilt angle. Hence, developing different aspects of the FSW technique operated by a
zero tilt angle could pave the way to make the FSW process feasible for industrial widespread use. In
this regard, the PI2 team supervised by Prof. Michel Guillot at Laval University is developing the
17
FSW technique at the right angle (RAFSW) [20, 24, 27-30]. It is demonstrated that using the
developed FSW technique has resulted in a 40 % reduction in the downward axial force compared to
the typical FSW processes at tilt-angle [20].
1.6.2 Effect of Tool Design on Downward Axial Force

Regarding the aforementioned issues, improving methods to reduce the axial force during the FSW
can lead to utilizing machines that have less load capacity. In addition, the FSW process under low
axial forces does not need robust fixtures to withstand process forces and be a repeatable process [8,
18]. In this connection, researchers have made some efforts. It is possible to reduce the needed axial
welding force to push the tool through the material by an effective tool design. Reducing the process
downward axial force is advantageous not only to reduce the force and energy costs but also to apply
higher traverse speeds [31]. For instance, it is clarified that using Flared-Triflute and A-skew pins, as
illustrated in Figure 13, leads to reduce the axial force by about 20% compared to conventional
threated pins. Moreover, it is demostrated that using Trivex pin results in an 18 to 25% reduction of
traversing forces and a 12% reduction in normal forces compared to the MX Triflute pin [1].
(a)
(b)
Figure 13: Schematic of: (a) Flared-Triflute tool, (b) A-skew tool developed by TWI from different views [1].
In another effort, it is demonstrated that using a special tool design can lead to obtaining a defect-free
weld using zero tilt angle. The effect of shoulder diameter, pin angle, and shoulder angle were studied.
These parameters are shown in Figure 14. The welds are done on 3.07 mm thick AA6061-T6 sheets
by a CNC machine tool and a designed tool to operate at the right angle. It is found that the shoulder
angle and diameter affect the axial force, mainly, while the pin angle has a minor effect. When the
shoulder angle, shoulder diameter, and pin angle were 4°, 12 mm, and 27°, respectively, the
downward axial force of welding reduces by 40% compared to the FSW process at tilt-angle, while
18
the mechanical properties of the welded joint were still acceptable. Figure 15 depicts the effect of
shoulder angle and diameter on the axial force in this research [20]. The Designed tool leads to reduce
the process costs due to some reasons, significantly. Firstly, it reduced the generated forces during
the process so that the needed features of the machines are less in terms of stiffness, sensitivity, and
clamping parts. Secondly, the needed flexibility of the machine is reduced as the process operated at
zero angle. This type of tool shows a depression in the joint and is more difficult to apply properly on
deeper welds.
Figure 14: Schematic of parameters of tool design [20].
Figure 15: Effect of shoulder angle and diameter on axial force [20].
1.6.3 Effect of Preheating on Axial Force

To decrease the amount of the generated axial force during the FSW process, some researchers have
focused on the effect of preheating on axial forces. It is shown that preheating of the workpiece leads
to the reduction of the axial forces during the FSW process. The samples were preheated at different
levels, and then welded at different travel speeds. The softening effect of the preheating of the
workpiece leads to the need for lower axial forces during the FSW process.
19
In another research, it is shown that the needed axial force during the FSW process can decrease by
resistance heating. In this case, flowing an electrical current between the workpiece and the welding
tool causes resistance heating of the workpiece. This leads to the reduction of the axial force. During
a conventional FSW method, the axial force is 20 kN to weld a 5 mm thick AA 5052 sample at 400
rpm and 600 mm/min. However, it is 5 kN when resistance heating is utilized. Figure 16 illustrates
the schematic of this method [32]. Although preheating seems a beneficial method to decrease the
axial force, implementation of such complex equipment is costly. This makes these methods
unreasonable from a cost perspective.
Figure 16: Schematic of resistance heating method [32].
1.6.4 Effect of plunge depth, tool rotation speed, and transverse speed on generated
axial force
Welding parameters have a considerable effect on the generated axial forces during the FSW process.
It is essential to have a deep knowledge regarding them to maintain the force at a low level, and
simultaneously, the tensile properties of the joint at an acceptable level. Then, it is possible to find an
optimized range of the generated forces to reach a sound, defect-free weld. In this regard, there is a
lot of research on the investigation into the effect of the tool rotation and transverse speeds on the
axial forces during the FSW process at different tilt angles [20, 26, 33, 34]. Whereas, there is lack of
such details for the FSW at the right angle.
Melendez et al. showed that the axial downward force decreases by increment of the tool traverse
speed. As, thermal diffusivity is restricted by welding speed, the ahead material would be cooler and
stronger by the increase of the speed. As a result, higher axial forces are needed to weld the AA6061-
T6 plates, as shown in Figure 17. Moreover, it is indicated that the increase in the tool rotational speed
leads to the reduction of axial forces during FSW of AA6061-T6 plates, as shown in Figure 18.
Because by the increment of rotational speed, the workpiece temperature would be higher which
means that the ahead material is softer. Thus, the FSW process generates less axial forces at a given
plunge depth [26]. A similar research has done on the FSW of 12 mm thick AA5083 plates by a CNC
20
milling machine at a 2° tilt angle. It was shown that the downward axial force decreases from 18 kN
to 17 kN by the increment of rotational speed. It is shown that by increasing of traverse speed, the
axial force increases from 17 kN to 18 kN, as shown in Figure 19 [33].
Figure 17: Axial force vs. tool rotational speed on 6.35 mm thick 6061-T6 and 2195-T6 friction stir welded at
800 rpm, 120 mm/min, and a 1° tilt angle [26].
Figure 18: Axial force vs. tool rotational speed on 6.35 mm thick 6061-T6 and 2195-T6 friction stir welded at
800 rpm, 120 mm/min, and a 1° tilt angle [26].
21
(a) (b)
Figure 19: Effect of (a) tool rotational speed and (b) traverse speed on the axial forces during FSW of AA
5083 [33].
In another research, the influence of the tool rotational and traverse speed on the downward axial
force has been studied. AA6061-T6 sheets of 3.07 mm thickness were joined by FSW at right angle
using a CNC machine tool. The tool had a shoulder angle, diameter, and a pin angle of 4°, 12 mm,
and 27°, respectively. As illustrated in Figure 20, the force is minimized at each travel speed along
with a specific rotational speed. The identified region of the tool rotational and traverse speed in the
figure is corresponding to the maximized UTS region [20].
Figure 20: Influence of tool rotational and traverse speed on the axial force during FSW process on 3.07 mm
thick AA6061-T6 at right angle [20].
22
Furthermore, the plunge depth of the FSW tool affects the downward axial force during the FSW
process, considerably [26]. As shown in Figure 21, the more the plunge depth, the higher the axial
downward force. Because at higher plunge depths, a larger plastic zone forms into a cooler material.
The cooler material has a higher resistance to the deformation. As a result, the more axial force would
be generated to mix and stir the material during the FSW process [26]. In another research, it is shown
that the plunge depth has a non-linear relation with the axial forces during the FSW at a tilt angle.
Because the stiffness of the workpiece is highly depended on the temperature. The use of a wider
range of the plunge depth is possible when the tilt angle increases [26]. While, the process is highly
sensitive to the plunge depth variations when a zero tilt angle is applied [4].
Figure 21: Axial force vs. plunge depth on a 6.35 mm thick 6061-T6 and 2195-T6 friction stir
welded at 800 rpm, 120 mm/min, and a 1° tilt angle [26].
1.7 Effect of tool design and process parameters on mechanical properties

of the FSW joints
The FSW tool design and process parameters play a key role in quality and final mechanical properties
of the joint. Regarding this issue, some efforts have been made on finding the relation between the
FSW them to optimize the process. The vast majority of the conducted research is done on the FSW
process at a tilt angle.
Figure 22 depicts the relationship between the tool design parameters and UTS of the welded joints
of 3.07 mm AA6061-T6 butt-joints welded by FSW at right angle. The identified region in the graph
23
corresponds with high UTS while the generated axial force is minimized. As seen in this figure, there
is a suitable range for shoulder angle and diameter to obtain sound, defect-free welds at low forces
[20].
Figure 22: Influence of the shoulder angle and diameter on the UTS of AA6061-T6 sheets welded by the
RAFSW [20].
Safeen et al. have made attempts to provide an overall understanding about the impact of the FSW
process parameters on the mechanical properties of FSW joints using response surface methodology
(RSM). For this purpose, 5 mm thick AA6061-T6 plates were welded applying 0.1 mm plunge depth
at different tool rotational and traverse speeds. Figure 23 depicts the effect of the rotational and
traverse speed on the UTS and hardness of the welded samples. The maximum UTS obtains when
FSW was done at a 3° tilt angle at the optimized rotational and traverse speeds of 1150 rpm and 70
mm/min, respectively. As seen in Figure 23, the UTS increases with the reduction of rotational and
traverse speeds, while the hardness increases to reach a peak and then decreases [35].
24
(a) (b)
Figure 23: Effect of (a) rotational speed and (b) traverse speed on UTS and hardness [35].
In another study, the effect of the FSW process parameters on the mechanical properties of aluminum
joints was investigated using RSM method. Table 4 demonstrates the optimized FSW process
parameters to obtain defect-free joints of AA1100 alloy. Figure 24 illustrates the effect of welding
parameters on the mechanical properties of AA1100 welded samples. Moreover, it was demonstrated
that the optimized rotational and traverse speed, and axial force for AA6061-T6 was 1178 rpm, 115
mm/min, and 8.2 kN, respectively, to obtain a weld with a UTS of 183.15 MPa [12].
Table 4: Optimized FSW process parameters for different alloys [12].
25
Figure 24: Effect of welding parameters on mechanical properties of AA1100 welded samples by FSW [12].
26
In another study, 4 mm thick AA 6061-T4 plates were welded by different FSW process parameters.
Then the optimized process parameters were found using analysis of variance (ANOVA). Figure 25
shows the related results. It was observed the optimum condition includes 920 rpm rotational speed,
78 mm/min traverse speed, and 7.2 KN axial force [36]. Rajakumar et al. have studied friction stir
welded AA 6061-T6 plates of 5 mm thickness using the RSM method. It was found that the optimum
range of process parameters are the rotational speed between 1155 to 1157 rpm, the traverse speed
between 84.51 to 84.67 mm/min, and the axial force of about 7.17 kN [37]. Elatharasan and Kumar
have performed some experiments on the FSW joints of 6mm thickness AA6061-T6 using ANOVA.
It was reported that the UTS increases up to a maximum amount by the increment of the axial force
and the tool rotational speed. By more increase of the axial force and rotational speed the UTS
reduces. The maximum UTS of 197.5 MPa was obtained by 1199 rpm rotational speed, 30 mm/min
traverse speed, and 9 kN axial force [38]. The reported results of this research contradict the
mentioned results by Safeen et al. It can originate from the differences in the tool design, tool tilt
angle and the range of applied process parameters [35, 38].
27
Figure 25: Effect of FSW process parameters on mechanical properties of AA6061-T4 welded samples by
FSW [36].
28
A study was conducted on the FSW welding of 2.1 thick AA6061-T651 sheets. It was found that a
quality weld obtains when the tool traverse speed, rotational speed, and plunge depth are more than
a certain amount. In this research, the tool tilt angle was 2.5°. The range of the tool plunge depth,
rotational speed, and traverse speed was from 1.5 to 2 mm, 800 to 1600 rpm, and 50 to 130 mm/min,
respectively. Figure 26 demonstrates the relation of axial force with tool rotation speed, traverse
speed, and plunge depth. To obtain a sound, defect-free weld, the safe range of these parameters is
indicated in this figure. Moreover, a domain of suitable FSW parameters was introduced as shown in
Figure 26. Inside this domain, the welded joint would be defect-free. The increase in traverse speed
and decrease of rotational speed leads to the increase in axial force. Besides, the increase in plunge
depth leads to an increase in the generated axial force, as shown in Figure 27. Then, for plunge depths
between 1.75 and 1.85, by the increment in plunge depth, the axial force increases, sharply. However,
the axial force increases by a gentle slope by more increment in plunge depth [39].
(a)
(b)
(c)
Figure 26: Relationship between axial force and (a) tool rotational speed, (b) traverse speed, and (c) plunge
depth with indication of the range of these parameters to obtain a defect-free weld [39].
29
Figure 27: The domain of parameters to obtain a defect-free weld [39].
Moreover, Kumar et al. have demonstrated the impact of the axial load on the tensile properties of
4.4 thick AA7020-T6 plates, which are friction stir welded at a 2° tilt angle. The initial plunge depth
of the tool was 3 mm. In this study, keeping the backing plate at a certain angle relative to the tool
axis in the welding direction causes the change of axial load due to the linear increment of interference
between the tool and the base material, as shown in Figure 28. This angle causes the increase in
plunge depth in the course of welding. Figure 29 demonstrates that the tensile strength increases by
increase of the axial load, firstly. Then it reaches a maximum and drops after. The results indicate
that the optimal downward axial load to obtain a sound, defect-free weld is 8.1 kN [34].
Figure 28: Schematic of the experiment [34].
30
Figure 29: The relation between the axial load and tensile strength [34].
1.8 Fatigue behavior of FSW joints made by different process parameters

and tool designs
Assessment of the fatigue behavior of a joint has crucial importance for structural applications.
Regarding this issue, there are a few studies for FSW joints. Costa et al. have investigated the fatigue
behavior of 4 mm thick AA 6082-T6 plates welded by the FSW at 2° tilt angle at butt joint
configuration. The fatigue test was done under constant and variable amplitude loading with the stress
ratio of R=0 (R =σmin/σmax) [40], and R= -1 [41] . It is found that the tunnel defect adversely affects
the fatigue behavior more than the shear lips. Avoiding such defects can enhance the fatigue behavior
of the FSW joints [41]. In another research, the fatigue behavior was investigated for 10 mm thick
AA7075-T651 plates that were welded at an 8° tilt angle. An ultrasonic fatigue test was used to study
high cycle fatigue behavior by applying loads at 20 Hz frequency. It was found that the fatigue crack
initiated from the TMAZ during a short fatigue life test. However, it initiated from the NZ during a
very high cycle fatigue test [42].
Ericsson and Sandstorm have studied the effect of the traverse speed on the fatigue behavior of
AA6082-T6 welded plates by the FSW. The fatigue test was conducted with a sinusoidal load-time
function with the stress ratio of 0.5, oscillation frequency in the interval of 9-15 Hz, and average
stress in the interval of 105-165 MPa. It showed when the traverse speed increases from 700 to 1400
mm/min, there was not a meaningful change in the fatigue properties of the joints. Whereas, there
was an improvement in the fatigue behavior of welded samples at 350 mm/min traverse speeds, as
shown in Figure 30 [43]. Lomolino et al. have made an effort to gather all the available fatigue data
for the FSW joints of aluminum alloys. Then, they conducted a statistical analysis to derive some
reference fatigue curves for FSW butt joints. They found that fatigue failure takes place within the
weld zone in most FSW samples. Moreover, the impact of R ratio on the fatigue behavior of FSW
31
joints of AA5083 was investigated. They suggested that mechanical machining leads to improvement
of fatigue properties of FSW joints [44].
Figure 30: Fatigue behavior of welded AA 6082-T6 samples at different traverse speeds [43].
1.9 Effect of joint fit-up variations on the quality of the FSW joints
The FSW process is sensitive to the workpiece variations such as a mismatch, gap, and misalignment.
Hence, it is important to identify the range of the disturbances tolerable by the process. The tolerable
disturbances means the range of joint fit-up variateions that do not lead to a considerable change in
the mechanical properties of the joint. Determination of the tolerable disturbances helps to ensure the
integrity of the FSW process. Moreover, the higher the tolerance of the varioations, the FSW process
would be more cost-effective and does not need expensive machines and high skilled labors. Also, it
helps to minimize the edge preparation for geometrical fit-up before welding [22, 45-47]. Figure 31
illustrates common mating variations in butt welding configuration [46].
Figure 31: Common mating variations in butt welding configuration; (a) gap, (b) mismatch, and (c)
misalignment [46].
Wanjara et al. have studied the gap tolerance at the butt welding configuration. A gap means a space
between the two sheets. The FSW process has been done by a robot on 3.18 mm thick AA6061-T6
plates at a fixed weld pitch of 0.48 mm/rev, weld pitch means the ratio of traverse speed to rotational
32
speed. It was found that if the gap size were over 0.5 mm, wormhole defects form. The maximum
tolerable gap size was 0.5 mm in this study [45].
Shultz et al. have evaluated the tolerable joint fit-up for a 5 mm thick 5083-H111 plates welded by
the FSW. The FSW was done at 1500 rpm rotational speed, 120 mm/min traverse speed, and different
travel angles of 1,3, and 5° by a robot, and at a 3° tilt angle by a CNC machine tool. Figure 32
demonstrates the effect of gap on the UTS and joint efficiency of the samples welded by an industrial
robot and a CNC milling machine using the 3° tilt angle. The results showed that for a given plunge
depth, the increase of the tool tilt angle results in the reduction of the effect of the gap. Thus, a 5° tilt
angle leads to higher UTS in comparison with the 1 and 3° tilt angles at a given amount of gap [22].
In another research, Shultz et al. have studied the tolerable joint fit-up during the FSW process to find
a gap compensation strategy. They conducted some experiments on 6 mm thick AA5083-H116 plates
welded by the FSW. The weld was done at the rotational and traverse speed of 1500 rpm and 120
mm/min, respectively, using an ABB robot operated under position control mode. According to the
conducted investigations, they recommended a strategy of sharing the control task between an
operator and a computer control system to compensate the gap effect. By this strategy, it is possible
to acquire higher gap tolerance compared to common methods [47].
Figure 32: Joint efficiency vs. UTS for an industrial robot and a CNC machine using a 3° tilt angle [22].
1.10 Effect of post weld heat treatment (PWHT) on mechanical properties

and fatigue behavior of FSW joints
Post weld heat treatment is an attempt to recover some portion of lost mechanical properties during
the welding process. There are some research in this regard. Zhang et al. have made an effort to boost
33
the mechanical properties of 6.5 mm thick AA2219-T6 plates welded by the FSW. The FSW was
done at the tilt angle, rotational, and traverse speed of 2.75°, 800 rpm, and 100, 400, and 800 mm/min,
respectively. Afterward, some samples were naturally aged for 10 days. Other samples were
artificially aged at 160°C for 16 h. It is found that the PWHT process enhances the hardness and
tensile properties of the FSW joints, considerably. Figure 33 illustrates the hardness improvement of
welded samples via PHWT. After PWHT, the strength of the joints made at high traverse speeds is
higher than the strength of the joints made by lower traverse speeds. It can be assigned to the
formation of more Al2Cu precipitates in the NZ at the heat-treated samples welded at high traverse
speeds [48].
Figure 33: (a) Hardness profile at different traverse speed for as-welded samples, effect of water-cooling
(WC) and post weld artificial aging (AA) on samples welded at: (b) 100 rpm, (c) 400 rpm, and (d) 800 rpm
[48].
In another study, the impact of the PWHT on the mechanical properties of the 3.17 mm thick
AA6061-T6 and AA6061-O plates welded by the FSW was studied. Each alloy were welded by the
FSW at rotational speed of 1000 and 1500 rpm and traverse speed of 150 and 400 mm/min. Welded
samples were solubilized at 803 K for 4 hours, and then water quenched and artificially aged at 443
K for 6 h, respectively. It is found that performed PWHT leads to the improvement in the hardness
distribution of both alloys. Figure 34 shows the hardness distribution of AA6061-T6 joints, before
34
and after the PWHT [49]. By the PWHT, the UTS of both alloys significantly improves even with
the presence of abnormal grain growth in the weld zone after the PWHT. Figure 35 demonstrates the
tensile test results of AA6061-T6 and AA6061-O welded samples at different rotational and traverse
speeds before and after the designed PWHT [49].
Figure 34: Hardness profile of AA-6061-T6 welded samples at different rotational and traverse speeds before
and after the PWHT [49].
(a)
(b)
Figure 35. Tensile test results of (a) AA6061-T6 and (b) both alloys welded at different rotational and traverse
speeds before and after the PWHT [49].
El-Danaf et al. Have conducted some experiments to restore the hardness and tensile properties of
friction stir welded AA6082-T651 samples. For this purpose, 6 mm thick AA6082-T651 samples
were friction stir welded at 850, 1040, and 1350 rpm rotational speed, and 90, 140, and 224 mm/min
traverse speed. Then, they were quenched in water, followed by artificially aging at 175°C for 5 and
35
or 12 h. The results showed that the proposed PWHTs could partially restore the softening effect of
the FSW process in the weld area. Moreover, the tensile properties recovered more by PWHT for 12
h than 5 h. Samples welded at lower speeds response notably to PWHT compared to samples welded
at higher speed, as shown in Table 5: The percentage of UTS improvement by PWHT [50]. This
behavior is assigned to the higher amount of available solutes in the matrix of low-speed samples to
re-precipitate the second phase during the PWHT process. Figure 36 demonstrates the effect of
PWHT on hardness and tensile properties of the samples [50].
Table 5: The percentage of UTS improvement by PWHT [50].
(a) (b)
Figure 36: Impact of PWHT on the hardness and UTS of the FSW joints welded at (a) different rotational
speed, (b) different traverse speeds [50].
Ipekoglu et al. studied the effect of PWHT on the enhancement of tensile properties of FSW butt-
welded AA7075-O and AA7075-T6 samples. It was found that the UTS of T6 joints were restored,
significantly, by solubilizing at 485°C for 4 h followed by quenching, and then artificial aging of the
joints at 140°C for 4 h [51]. Sivaraj et al. have evaluated the effect of PWHT on friction stir welded
12 mm thick AA7075-T651 samples. It was revealed that artificial aging PWHT for 24 h at 120°C
leads to a reduction in the tensile properties of the joints compared to as-weld samples. It was found
that solubilizing at 480°C for 1 h followed by artificial aging at 120 °C for 24 h enhances tensile
properties and hardness of welded samples, as shown in Figure 37 [52].
36
Figure 37: Hardness profile of as-weld samples (AW), artificial aged (AA), and solubilized followed by
quenching and artificial aging (STA) [52].
1.11 Applications of FSW

The FSW is utilized in various applications from transportation to aerospace and maritime industries
[1]. From 1998 a new series of investigation on using FSW in automotive industries were begun by
the contribution of TWI with other companies such as BMW, Ford, Daimler-Chrysler, General
Motors, Land Rover, and Volvo. These attempts paved the way to implement FSW in manufacturing
automobiles. For instance, the FSW has been used by Ford to make the lap joint from the stamping
and extrusions for the central tunnel of Ford GT, as shown in Figure 38. The FSW is used by Mazda
for the rear door structure of RX-8, by Audi for tailored blank aluminum plates in R8 high-
performance sports car, by Honda to produce a lightweight engine cradle in 2011 Honda Accord, and
by Mercedes for central tunnel and also for joining of extruded floor panels in 2012 SL model [4].
Another industry that is taking advantage of using the FSW is the aerospace industry since it is a
repeatable and reliable joining process that can be applied to lightweight alloys [4]. Some of the
widely used alloys in the aerospace industry are not weldable or it is difficult to weld them by
conventional fusion welding methods. Therefore, utilizing the FSW can be a great solution for this
issue [1]. Furthermore, implementing FSW in the maritime industry has been noted for some reasons.
As an instance, the heat input in the FSW process is less than fusion welding techniques. This results
in less residual stresses and distortion in the welded part. Hence, it leads to reduce the need for post-
process and distortion mitigation technique that totally, cause to cut the cost of production [4].
37
Figure 38: Friction stir welded central tunnel assembly of the Ford GT which is made up of aluminum
stampings and extrusions [4].
Nowadays, lap joints are widely utilized in the transportation industry to assemble different parts and
products. For instance, it is used for the assembly of aircraft skin and stringers, ship decks, railway
tankers, good wagons, supporting frame of pressure vessels, and shape frame of structures. These
joints are mainly made by fusion welding techniques such as MIG and laser welding or mechanical
assembly techniques such as riveting. Figure 39 illustrates a structure including lap joint
configuration. It is mainly assembled by riveting. Fusion welding techniques such as MIG welding
cause a high level of residual stress and distortion in the welded part. Moreover, it causes the
formation of more severe defects than the FSW defects in the welded area such as porosities. Besides,
the HAZ area is normally bigger than the HAZ in the weld made by the FSW. These issues cause low
dimensional accuracy of the welded part and considerable degradation of the mechanical and
corrosion properties of the joint. Mechanical assembly by riveting causes the increment of the weight
of the structure. Besides, the rivet holes act as potential sites for crack initiation and final fracture of
the structure, especially under dynamic loads [53-59]. To address such issues, the FSW could be a
promising assembly technique. Two configurations for the lap joint are shown in Figure 40 [60].
38
Figure 39: A structure consisting lap joint that is mainly assembled by riveting.
Figure 40: Lap joint configurations: (a) the upper sheet on the advancing side, (b) the lower sheet on the
advancing side [60].
One of the main fields of the widespread use of FSW could be transportation industries. For instance,
using that to assemble truck panels, railway cars, electric buses, trains, bridge decks, ships, and
airplanes [61-63]. Figure 41 illustrates some hollow extrusion aluminum profiles for railway cars
which could be produced by the FSW technique [61, 62]. Evaluation of the mechanical and fatigue
performance of these structures is of great importance to find the best welding process parameters.
Figure 42 demonstrates the failure of such structures under the dynamic loading [61].
In addition to the unique benefits of utilizing the FSW process as a joining technique instead of the
other conventional welding methods, there is a growing interest to take advantage of friction stir
technology in other advanced applications. According to FSW technique, some new technologies are
developed, or have emerged recently or are in the developing process.
39
Figure 41: Aluminum extruded profile to make railway cars body by FSW (1: before welding, 2: after
welding) [61, 62].
Figure 42: Failure position and stress prediction around the joint [61].
Figure 43 depicts an overview of friction stir-based technologies introduced during the timeline of
the past 26 years up to now [64]. As seen, there is ample room to implement friction stir technology
in different applications. For instance, rapid tooling to produce complex components by a
combination of friction stir spot welding and machining is reported [65]. It is reported that additive
manufacturing to make simple layer-by-layer components could be implemented by friction surfacing
[66]. Therefore, the FSW process has been an inspirational source to invent new manufacturing
techniques by its combination with other production methods.
40
Figure 43: An overview on categorized friction stir based technologies [64].
41
Problems, Objectives, and Methodology
2.1 Problem statement
2.1.1 FSW
Despite the short history of the FSW technique, it has affected the welding community, considerably,
in terms of not only volume of research but also its implementation in several industrial applications.
It provides users with many benefits such as higher strength, increased fatigue life, lower distortion,
less residual stress, less sensitivity to corrosion, and defect-free joints in comparison with arc welded
joints [8, 19]. At a glance, it is estimated that if just 10 percent of the US commercial welded joints
were replaced by the FSW joints, it can make around $5 billion/year financial benefit to the US
industry. Besides, it leads to around 1.3×1013 Btu/year energy saving, and 500 million lb/year
reduction in the greenhouse gas emission [8]. However, extensive commercial implementation of this
welding technique instead of other conventional techniques has remained a controversial issue for
most industries and applications, yet. Machine costs and royalties are the major barriers against the
widespread implementation of the technique. Moreover, there are a lot of uncertainties and scientific
gaps, lack of process specifications, and lack of design guidelines which act as barriers [19]. Thus,
development of standards and process specifications, design guidelines, and further innovations and
the developments of the process to introduce low-cost FSW techniques is of vital importance to make
the process affordable and push the industry to go beyond its comfort zone and take advantage of this
technology [8, 19].
2.1.2 RAFSW
Conducted research up to now mainly demonstrates that a non-zero tool tilt angle should be applied
during the FSW process to accomplish a sound defect-free weld [18]. 4 or 5-axis machines are needed
to do the FSW at tilt angle, while FSW at zero angle can be done by 3-axis machines. Moreover,
lower levels of axial forces generates during the FSW at the right angle. Considering these issues,
developing FSW techniques to apply zero tilt angle can make a great change in the needed capabilities
of the FSW machines so that the technique would be more cost-effective for industrial use.
In this regard, a low-cost friction stir welding technique at right angle (RAFSW) has been recently
developed by PI2/REGAL team at Laval University under supervision of Prof. Michel Guillot [29].
The RAFSW technique can be employed on low-cost 3-axis CNC machining centers without any
need of prior modification of the machine. The RAFSW technique is capable of making sound, defect-
free welds at a zero tilt angle with low axial forces during welding compared to common FSW
techniques. Thus, this method not only works with 3-axis machines instead of 4- or 5-axis machines,
42
but also does not need sturdy, stiff and high capacity expensive equipment. Moreover, clamping and
fixturing can be easily done by existing vises and clamps of the CNC machining center since the
required holding force is lower [29]. Additionally, as this technique does not need any modification
of the CNC machine, the same machine can be used for the machining of the parts before the welding
or for the finishing process after the welding. The price of such low capacity 3-axis CNC machine is
much lower than common FSW machines in the current industry. Furthermore, the reduced cost of
clamping and fixturing besides the possibility of the use of the same machine for machining at the
same series of actions, make the process even more cost-effective. The RAFSW technique has been
introduced, recently. There is lots of room to develop different aspects of the technique for various
applications and to make it reliable for industrial use.
2.1.3 Tool design and process parameters

In the FSW process, the tool design and process parameters affect the generated axial forces and the
physical and mechanical properties of the joints. There is a good deal of reported data regarding this
issue for the aluminum joints made by the FSW process [21, 26, 33, 34, 37-39, 67-69]. The reported
research is mainly done on the FSW at tilt angle. Thus, there are serious deficiencies in data regarding
the FSW at right angle.
As the tool plunge depth, rotation and transverse speed directly control the mixing and stirring process
during the FSW, they control the amount of generated force which could be the reflection of resistance
of material against deformation. Moreover, the tool design has a great impact on the mixing and
stirring process. Therefore, understanding the combining effect of the process parameters and a
certain tool design on the generated axial forces and the mechanical properties of the joint is essential.
The combining effect of these parameters on the axial forces and the quality of the joint raise from
the reaction of each material to a certain tool design during the friction stirring process. Indeed, the
friction characteristics of the material and the stirring mechanisms provided by a certain tool design
could be affected by tool plunge depth, rotation and transverse speeds, significantly. The vast majority
of the available information on the effect of process parameters and tool design are for FSW at a tilt
angle. Moreover, a great deal of reported research are done at low traverse speeds so that they are not
feasible for commercial use. There is a severe lack of data on the effect of these parameters for the
RAFSW process, especially at high traverse speeds so that be comparable with MIG best welding
speeds.
In addition, the vast majority of available data are for FSW processes at high axial forces. The amount
of axial forces has a complex relationship with the mixing and stirring mechanisms during the FSW
process. While there is a lack of scientific data onto the effect of welding parameters of RAFSW
43
process that has lower axial forces. Therefore, it is essential to conduct deep research on the
relationship between the RAFSW process parameters and tool design with mechanical properties of
the welded joints, especially at low axial forces and high tool transverse speeds. Moreover, it is crucial
to identify the efficient working window of the process parameters and tool design for a given
application [8, 18, 19]. Additionally, every effort to make a technique commercial relies on the
repeatability and robustness of the process. Hence, it is important to determine the tolerable amounts
of disturbances during the process.
2.1.4 PWHT
Another barrier against the widespread implementation of the FSW in the industry is the fact that
there are a lot of uncertainties, scientific gaps, and lack of guidelines regarding other aspects of this
process like post-weld heat treatment (PWHT) of FSW joints [19]. It is confirmed that the PWHT
could greatly enhance the mechanical properties of the FSW joints [48-50], but the reported research
is limited. There is a lack of research on the effect of the PWHT on the FSW joints, especially for
RAFSW samples. Therefore, it is essential to shedding light on the impact of various PWHTs
processes on the optimized welded samples by the RAFSW technique to make it reliable for potential
industrial users. Since the developed RAFSW technique has its certain features in terms of tool
design, involved forces, and optimized process parameters used for welding [70], it is necessary to
investigate and characterize the effect of different natural aging, artificial aging, and solubilizing
processes on the joints made by this technique.
2.1.5 Applications
Implementing the FSW in various industries is going to continue with an accelerated pace due to its
unique and prominent advantages over conventional fusion welding methods [19]. The FSW joints
have higher fatigue life than fusion-welded joints such as MIG welded joints. Furthermore, less
generated heat, absence of filler material, fine equiaxed microstructure in the nugget zone, and low
amount of porosity generation during FSW are some other benefits of the FSW compared to
conventional welding methods. Implement of the FSW could lead to the higher mechanical and
corrosion properties of the joints. Additionally, it could save a considerable amount of time and
money by increment of the lifetime of the components, weight reduction, and elimination of the post
processes such as shot pinning and machining. Furthermore, assembly of some components in the lap
configuration is not possible by the conventional fusion welding. Because the amount of the heat
input is high. However, the FSW could be a likely solution to them as it generates low levels of heat
input. In this regard, there are only a few reported research [53-59, 71]. These reported researches are
mainly on FSW at a tilt angle and they are done at low welding transverse speeds. Thus, there is
44
dearth of knowledge regarding the implementation of the FSW, specifically RAFSW, for different
components, and alloys at different configurations.
In addition, applying the FSW process for assembly of the components with complicated shapes has
remained a practical gap, yet. Therefore, exploring the well-suited welding parameters, the proper
joint design, and the proper fixturing for them make one able to taking advantage of the FSW in a
wider range of applications. One of the main fields of the widespread use of FSW could be
transportation industries. For instance, using that to assemble the parts with different geometrical
features for truck panels, railway cars, electric buses, trains, bridge decks, ships, and airplanes [61-
63].
Indeed, there is only a few researches available on the investigation into the fatigue behavior of the
standard FSW joints mainly made at a tilt angle. Unlike the FSW at a tilt angle, FSW at the right
angle does not make thin the areas around the joints, so, it could lead to enhancement of the fatigue
behavior of the joints. As there are no remnant grooved weld edges in welded samples by FSW at the
right angle, it can cause a longer time for crack initiation under loading conditions. Indeed,
widespread implementation of FSW instead of other conventional welding methods needs some firm
scientific data to show that this assembly is reliable enough under the application condition. In this
regard, knowledge of fatigue behavior of the welded parts is of vital importance, especially for
complex-shaped components under the real dynamic loading condition. Thus, conducting some
surveys to overcome the dearth of knowledge regarding this issue could be helpful to validate the
reliability of this technique for the industrial use.
Finally, there is an ample room to implement friction stir technology in different applications. The
FSW process has been an inspirational source to invent new manufacturing techniques by its
combination with other production methods. Taking advantages of RAFSW technique by CNC
machines, the RAFSW can open the door to introduce new manufacturing techniques.
45
2.2 Hypothesis
FSW is a new welding technology that has prominent advantages over fusion welding techniques
such as MIG welding.
2.3 Objectives of the thesis

In this thesis, the main goal is to continue to develop the RAFSW technique operated on common
inexpensive 3-axis CNC machines to make it reliable and affordable for potential industrial users. To
provide an overview, there are five or six FSW machines available in the Province of Quebec, while
there are many thousands of CNC milling machines capable of operating the proposed RAFSW
technique.
To pave the way for potential users to take advantage of this technique, this thesis will focus on
developing the scientific and technical knowledge to operate the technique by optimized process
parameters and tool design. The results will provide researchers and users with a guideline that can
be utilized for various joint designs and applications. Therefore, this thesis starts from the final results
of the previous Ph.D. thesis [29] of the team. The initial goal is to go beyond that research and gain
deeper insight through what has been done, previously, at higher traverse speeds. To put step further,
the recovery of the joint properties will be studied by the PWHT techniques. Afterwards, the RAFSW
method will be developed for lap joint configurations by optimizing the tool design and process
parameters. Then, the process and the tool design will be adapted to operate on force control mode.
Afterwards, the process will be implemented to assemble large truck side panels using a big router
CNC machine while minimizing the need for clamping to industrialize the process. Therefore, the
following topics are the main contribution of the current thesis:
• Investigating the effect of process parameters of RAFSW on the downward axial force and
mechanical properties of the butt-welded joints
• Study the effect of PWHT process on the RAFSW butt-joints
• Design and optimization of the tool specifications and process parameters for lap joints
welded by RAFSW operated on a CNC milling machine
• Adapting the process parameters, tool design, and control mode to assemble large truck side
panels using a big router CNC machine
46
2.4 General Methodology
Nowadays, Taguchi method is extensively used to optimize manufacturing processes [72]. Taguchi
design of experiments (DOE) is very effective to minimize the needed number of experiments to
investigate the effect of input process parameters on the output results. After conducting the
experiments based on the DOEs, artificial neural network (ANN) modeling, instead of analysis of
variances (ANOVA), has been used to model complex relationships between inputs and outputs, in
this study [72]. Taking advantage of Taguchi method along with high predication capability of ANN,
one would be able to predict the relationship between inputs and outputs through minimized number
of experiments.
In this research, feed-forward neural networks with backpropagation algorithm has been used to
model the relationship between the tool geometry and the process parameters on the downward axial
force and the fracture force of the joints. Several factors affect the prediction accuracy of the
backpropagation neural network models such as architecture of the network, momentum coefficient,
and learning rate of the model [73, 74]. An ANN model with too small architecture can results in
insufficient degree of freedom; and too large network causes to over fit the data. Thus, there is an
optimized number of neurons and hidden layers to have a reliable ANN model [75]. The factors to
compare the different architectures are root-mean squared error (RMSE), maximum error, mean
relative error (MRE), and mean absolute error (MAE) [76]. Furthermore, the relationship between
the experimental data and predicted values by the developed ANN models has been studied by
calculating the amount of correlation coefficient (R2) using linear regression analysis. When the
correlation coefficient is close to one, it indicates a close relationship between the experimental data
and predicted data by the developed ANN models [73, 76, 77].
In this research, the ANN modeling is done using a software developed by Prof. Michel Guillot. The
author does not have any contribution in the development of this software and have only used it to
analyze the data obtained from the experiments. The input parameters to study are the tool design
parameters and process parameters. The outputs are downward axial force and either the strength or
the failure force of the joint. Some details regarding the theory of the ANN modeling and a case study
on the developed ANN models in this research is provided in appendix.
The RAFSW technique has been developed to make welds using an inexpensive 3-axis CNC machine,
Fryer MC15 machine in position and force control modes. Moreover, the same RAFSW technique
has been implemented on a big CNC router that was less stiff than the Fryer MC15 machine.
The material to make welds is extruded plates and sheets of AA6061-T6 of 6.35 and 1.6 mm-thick.
The samples has been welded in butt joint and lap joint configurations. Finally, the technique has
been implemented to assemble large delta-shaped truck side panels of AA6061-T6. The robustness
47
of the developed technique has been studied by investigating the effect of joint fit-up disturbances.
Additionally, the effect of PWHT has been studied by applying various artificial aging and
solubilizing processes.
To study the quality and reliability of the welded parts, various characterization techniques has been
used. The conducted tests are uniaxial tensile test, Vickers micro-harness test, metallography and
microscopic observations, and fatigue test. Moreover, the strength of the truck panel samples has been
validated using finite element simulations followed by experiments. Finite element simulations by
NX-NASTRAN software has been used to benchmark the level of load that can be sustained while
the experimental tests have been done to identify the real strength of the parts.
48
Development of Friction Stir Welding
Technique at Right Angle (RAFSW) Applied on Butt-
Joint of AA6061-T6 Aluminum Alloy
Mahboubeh Momeni, Michel Guillot*
Laval University, Department of Mechanical Engineering, Aluminum Research Center – REGAL,

PI2 research team, 1065, Ave de la Médecine, Quebec, Canada, G1V 0A6
* Correspondence: mguillot@gmc.ulaval.ca; Tel.: +1-418-988-6549, +1-418-656-3343
Keywords: Friction stir welding; Axial forces; Tensile properties; Taguchi design of experiments;
Neural network modeling
This Chapter has been published in: The International Journal of Advanced Manufacturing
Technology, Vol 99, 3077–3089 (2018)
https://doi-org.acces.bibl.ulaval.ca/10.1007/s00170-018-2672-8
49
3.1 Résumé
Le soudage par friction-malaxage (FSW) est un processus d'assemblage à l'état solide, qui présente
des avantages importants par rapport aux méthodes de soudage par fusion courantes comme le
soudage MIG. Cependant, les coûts d’équipement et les redevances constituent les principaux
obstacles à l’utilisation généralisée de cette technique dans l’industrie actuelle. Ainsi, le
développement de techniques FSW à faible coût pourrait ouvrir la voie à des utilisateurs industriels
potentiels. Dans cette recherche, nous allons caractériser et optimiser une technique FSW à angle
droit récemment développée (RAFSW) à l'aide de machines-outils CNC 3 axes à faible coût. À cette
fin, une étude complète a été effectuée pour obtenir un aperçu approfondi de l'effet des paramètres du
procédé RAFSW sur les forces axiales générées et les propriétés de traction des barres plates soudées
bout à bout AA6061-T6 de 6,35 mm d'épaisseur. La planification des expériences et la modélisation
des résultats ont utilisé respectivement la méthode de conception des expériences de Taguchi et la
modélisation des réseaux de neurones artificiels. À partir de ces modèles, une fenêtre de travail
efficace des paramètres de processus a été établie afin que les machines CNC avec une capacité et
une rigidité minimisées puissent produire des soudures de haute qualité de manière répétée. Après
vieillissement naturel, la résistance ultime à la traction a atteint un maximum de 253 MPa et reste
supérieure à 247 MPa dans la fenêtre de travail. De plus, la robustesse de la technique et l'effet du
vieillissement artificiel sur les échantillons les plus forts ont été évalués. Nous avons constaté que la
résistance à la traction ultime des échantillons vieillis artificiellement atteignait 300 MPa. Enfin, des
images de microstructure et des mesures de microdureté des échantillons soudés les plus solides
complètent l'analyse. Les résultats de cette recherche ont un grand impact sur la fiabilité de la
technique RAFSW récemment proposée pour les utilisateurs industriels.
3.2 Abstract
Friction stir welding (FSW) is a solid-state joining process, which has prominent advantages over
common fusion welding methods like MIG welding. However, equipment costs and royalties act as
the main barriers to the widespread use of this technique in today’s industry. Thus, developing low-
cost FSW techniques could pave the way for potential industrial users. In this research, we are going
to characterize and optimize a recently developed FSW technique at right angle (RAFSW) using low-
cost 3-axis CNC machine tools. To this end, a full study was done to gain a deep insight through the
effect of RAFSW process parameters on generated axial forces and tensile properties of AA6061-T6
butt-welded flat bars of 6.35mm thick. Experiment planning and modeling of the results used Taguchi
design of experiments method and artificial neural network modeling respectively. From these
models, an efficient working window of process parameters has been established so that CNC
50
machines with minimized capacity and stiffness can produce high quality welds repeatedly. After
natural aging, the ultimate tensile strength reached a maximum of 253 MPa and stays above 247 MPa
within the working window. Moreover, the robustness of the technique and the effect of artificial
aging on the strongest samples have been evaluated. We found that the ultimate tensile strength of
artificially aged samples attained 300 MPa. Finally, microstructure images and micro-hardness
measurements of the strongest welded samples complete the analysis. The outcomes of this research
have a great impact on making the recently proposed RAFSW technique reliable for industrial users.
3.3 Introduction
In recent decades, friction stir welding (FSW) emerged as one of the most significant developments
among metal joining processes [1, 3]. Indeed, FSW is a solid-state metal joining process that provides
users with many advantages over fusion welding techniques such as higher energy efficiency,
versatility, environmental friendliness, and the capacity to join non-weldable and dissimilar alloys.
Moreover, FSW joints demonstrate higher tensile strength and fatigue life, lower distortion, less
residual stress, less sensitivity to corrosion, and porosity-free joints in comparison with conventional
fusion welded joints [1, 2, 4, 8, 19]. In spite of prominent advantages of FSW over common fusion
welding techniques, equipment costs and royalties are the major barriers against the widespread
implementation of FSW [8, 18, 19].
During the FSW process, a non-consumable rotating tool plunges into a workpiece and traverse along
the joint line. The main process parameters of the FSW are tool tilt angle with respect to the sample
surface, tool plunge depth, tool rotational speed, and traverse speed [1, 2]. These process parameters
along with tool design directly control the mixing and stirring process during the FSW. Thus, they
affect the amount of forces generated during material deformation. Moreover, the quality of the weld
is the direct result of an effective mixing and stirring process [1]. Therefore, understanding the
combined effect of these parameters on generated forces and quality of the weld is of crucial
importance to set properly these parameters. Although, there is a good deal of reported research
available regarding the effect of FSW parameters on generated force and mechanical properties of
the welded aluminum plates at tilt angle [26, 34, 37-39, 68, 69], there is a severe lack of information
available regarding the quality, defect-free FSW joints welded at right angle. Moreover, in the vast
majority of the reported research, FSW joints are welded at such low traverse speeds that they are not
economical for most industrial applications.
Indeed, conducted research mainly demonstrates that a non-zero tool tilt angle should be applied to
accomplish a sound defect-free weld [18]. In addition, other authors reported that the higher the tilt
angles, the greater the axial force generated by the FSW process [25, 26]. As a result, this process
51
often uses specialized sturdy FSW machines capable of withstanding very large forces with 4 or 5-
axes to produce the tilt angle. Accordingly, these machines are very expensive and their use limited
to FSW only.
To make the FSW process cost-effective, PI2/REGAL team at Laval University has recently
developed a low-cost FSW technique at right angle, called RAFSW [20, 24, 27-29]. In this technique,
commonly available 3-axis CNC machine tools used for machining can also make RAFSW without
modifications and at low equipment cost. However, to make it applicable for industrial use, a deeper
study of the RAFSW technique in a wider range of process parameters is necessary.
Thus, the main objective of the present paper is to apply the RAFSW technique and study the effect
of multiple process parameters on maximizing the weld strength while minimizing the force required
during welding. From this study, we establish a working window of parameter settings for maximum
performance and reliability. Moreover, some investigations will be done on the effect of post weld
artificial aging on tensile properties, robustness of the RAFSW process, microstructure and micro-
hardness of the best welded sample.
3.4 Experimental procedure

In this study, extruded AA6061-T6 flat bars, 6.35 mm thick and 76.2 mm wide, are butt-welded. For
the welding tests, two bars are cut 250 mm long using a power hacksaw. These flat bars are clamped
onto a rigid plate installed on a 3-axis Kistler 9265B dynamometer as shown in Figure 44. The
dynamometer itself is bolted to the table of a Fryer MC-15 3-axis CNC machining center. This
machine has a 25 HP spindle using CAT40 tool holders and rotating to a maximum of 8000 rpm.
Axis peak trust is 15 kN but the controller triggers an alarm between 9 and 10 kN. Even though the
machine accuracy is stated at +/- 0.004 mm, under load this may differ slightly.
For RAFSW, a specially designed tool is mounted into a long reach tool holder as illustrated in Figure
44. This tool has a shoulder diameter of 15 mm and a pin length of about 6 mm [20, 27].
Single pass butt welds are done along the extruded direction of the bars. Although most of the tests
have been carried out with an accurate positioning of the bars, to evaluate the robustness of the
process, we also added some tests with slight positioning errors like gap, mismatch, and
misalignment. Gap means the opening distance between the bars prior to welding; mismatch refers to
the height difference in the top surface of the plates; and misalignment means that the FSW tool is
not oriented exactly in the interface centerline between the abutted bars. The mentioned variations
are made either by machining the bars or by using shims as spacers.
52
3.4.1 Material properties
The AA6061-T6 alloy has a nominal chemical composition of 0.99 wt.% Mg, 0.58 wt.% Si, 0.20
wt.% Cu, 0.35 wt.% Fe, 0.12 wt.% Cr, 0.04 wt.% Mn and the rest is aluminum. The ultimate tensile
strength and ductility of the base material are 355 MPa and 13.6%, respectively.
FSW tool
Clamps
Abutted
Aluminum
edge
bars
Plate
(a)
Dynamo-meter
(b)
Figure 44: (a) FSW tool and tool holder, (b) Welding set-up including the extruded bars, the clamping system
and the Kistler 3-axis dynamometer.
3.4.2 Quality evaluation

To evaluate the tensile properties of the joints, tensile samples were machined from the weld coupons
according to the American Society for Testing of Materials (ASTM E8M-04) standard, as shown in
Figure 45. All samples have been naturally aged for at least 7 days at room temperature prior to
testing. Uniaxial tensile tests were conducted using a hydraulic testing machine equipped with a load
cell of 44.5 kN calibrated to +/-0.08% under a cross head speed of 1 mm/min. A high-resolution
Epsilon 25 mm extensometer, was used during uniaxial tensile tests.
Figure 45: Schematic of tensile test samples.
3.4.3 Metallography
The weld cross section is prepared using standard metallographic practice. The samples are cut,
mounted in phenolic resin and polished to 2000 grits papers using an automatic Buehler polishing
machine. For etching of AA6061-T6 welded samples Weck’s reagent was found to work, effectively
53
[78]. Afterwards, micrographs are taken using a Nikon K100 optical metallographic microscope.
Furthermore, macro-etching process was done by a 75 ml HCl + 25 ml HNO3 + 5 ml HF solution
diluted with 25% distilled water.
3.4.4 Micro-hardness
Micro-hardness measurements were made across the weld line with a load of 300 g and a dwell time
of 15 s using a Buehler Vickers micro-harness tester and applying ASTM E384 standard.
3.4.5 Post weld heat treatment

Finally, to study the effect of post weld heat treatment on the recovery of tensile properties, the
strongest samples were subjected to standard artificial aging process at 160 °C for 18 h in a calibrated
PYRADIA furnace.
All metallographic, hardness, and post welding heat treatment (PWHT) investigations were done
using the joints welded at optimized condition, which was tool plunge depth, rotational speed, and
traverse speed of 6.15 mm, 3500 rpm, and 540 mm/min, respectively.
3.5 Design of experiments and modeling

3.5.1 Design of experiments
Nowadays, Taguchi method is extensively used to optimize manufacturing processes [72]. Taguchi
design of experiments (DOE) is very effective to minimize the needed number of experiments to
investigate the effect of input process parameters on the output results such as ultimate strength (UTS)
and plunging force of FSW. After conducting the experiments based on the DOEs, artificial neural
network (ANN) were used to model complex relationships between inputs and outputs [72]. Taking
advantage of Taguchi method along with high predication capability of ANN, one would be able to
predict the relationship between inputs and outputs through minimized number of experiments. In
this research we have implemented a technique that consists of a sequential series of design of
experiments and neural network modeling. Each series of experimental tests was followed by training
of an ANN model, and then the next series of experiments was designed based on this developed
ANN model. Hence, the next experimental data set not only is a set of confirmation tests for the
previous ANN model, but also will be added to the training data for the next step to train the next
ANN model by both new and last datasets.
In this research, the first DOE is an L8 orthogonal array used to explore the effect of RAFSW process
parameters on the generated axial force during welding, F, and the ultimate tensile strength, UTS, of
the joints. The investigated process parameters are the welding transverse speed, V, the tool rotational
54
speed, w, and the tool plunge depth, P.D. The experimental results of the first DOE, shown in Table
6, was used to train the first ANN model. Afterwards, based on this developed ANN model, a second
DOE, an L4, was designed to explore more around the optimal region identified by the ANN model
(see Table 7). Then, the experimental results of both, the first and second, DOEs were used to retrain
a second ANN model. Again, based on the estimation of the second ANN model, a third DOE, shown
in Table 8, was designed to refine the optimal region of the second ANN model. Moreover, another
set of experiments was done to enrich the ANN model with data shown in Table 9. Based on the next
retrained ANN model, it was found that the highest tensile strength is achieved when the tool plunge
depth, rotational speed, and traverse speeds are 6.15 mm, 3500 rpm, and 540 mm/min, respectively.
Based on this result, another set of experiments depicted in Table 10, was designed to explore the
effect of tool plunge depth on force and tensile properties when the tool rotational and traverse speeds
are at optimal levels of 3500 rpm, and 540 mm/min. These experimental results have confirmed that
the highest UTS is accomplished at the suggested optimal conditions by the previous ANN model.
Again, this set of experiments was added to the training data of the ANN model to develop a more
precise model around the optimal area. Therefore, to study the force and UTS of the RAFSW joints,
the last ANN model before confirmation was trained using data of 26 experiments. Then, four
confirmation tests, shown in Table 11, demonstrate the accuracy of the models for both axial force
and UTS. Finally, confirmation dataset was also used to enrich the final ANN model retrained using
30 experimental data. From the final ANN model, welding at tool plunge depth, rotational speed, and
traverse speed of 6.15 mm, 3450 to 3550 rpm, and 535 to 545 mm/min yields the highest tensile
strength while the force is lower than 6 kN.
Table 6: The first DOE.
L8 P.D. (mm) w (rpm) V (mm/min) F (kN) UTS (MPa) El % Y (MPa)

1 5.75 2500 500 3.4 143 1.8 130
2 5.75 3500 300 3.7 210 6.9 129
3 5.9 2500 300 5.9 208 7 128
4 5.9 3500 500 3.5 142 2 135
5 6.05 2500 500 5.9 222 9.5 128
6 6.05 3500 300 5.8 208 7.8 129
7 6.15 2500 300 5.7 217 8 128
8 6.15 3500 500 5.7 248 12 142
55
Table 7: The second DOE
L4 P.D. (mm) w (rpm) V (mm/min) F (kN) UTS (MPa) El % Y (MPa)

1 6.1 3900 600 5.7 224 9 135
2 6.1 3750 560 5.1 237 8.5 132
3 6.2 3750 560 7 242 8.6 130
4 6.2 3900 600 6.7 242 8.9 134
Table 8: The third DOE (P.D.=6.15 mm)
L4 w (rpm) V (mm/min) F (kN) UTS (MPa) El % Y (MPa)

1 3800 560 6.3 245 9.1 135
2 3800 600 6 234 8 145
3 4000 560 6.3 244 10.4 135
4 4000 600 5.9 237 8.3 131
Table 9: 4th dataset of experiments to enrich the ANN model.
P.D. (mm) w (rpm) V (mm/min) F (kN) UTS (MPa)

1 6.1 3650 600 5.8 248
2 6.12 3300 530 4.5 218
3 6.2 3700 650 3.6 138
4 6.3 3300 650 5.4 248
Table 10: Some experiments to explore about the effect of plunge depth on force and UTS for joints welded at
optimal tool rotational and traverse speeds.
P.D. (mm) w (rpm) V (mm/min) F (kN) UTS (MPa)

1 5.95 3500 540 3.8 142
2 6 3500 540 4 146
3 6.05 3500 540 4.1 154
4 6.1 3500 540 5 242
5 6.15 3500 540 5.9 253
6 6.2 3500 540 6.3 249
Table 11: Confirming tests for the ANN model.
P.D. w V F UTS El Y ANN-UTS Er-UTS ANN-F Er-F

(mm) (rpm) (mm/min) (kN) (MPa) % (MPa) (MPa) (kN)
1 6.1 3550 560 5.2 240 7.8 127 248.3 -3.5% 5.44 -4.6%
2 6.15 3900 590 6.1 242 8.3 134 240.8 0.5% 6.08 0.3%
3 6.25 3900 600 7.4 238 7.6 130 241.6 -1.5% 6.91 6.6%
4 6.3 3700 530 6.1 243 8.1 132 244.9 -0.8% 6.09 0.2%
56
In a second phase, the robustness of the RAFSW process is investigated using an L9 orthogonal array
designed to evaluate the effect of joint fit-up disturbances on the tensile strength of the welds, as
shown in Table 12. For this series of experiments, the optimal tool rotational and traverse speeds of
3500 rpm, and 540 mm/min were employed. The ANN model was trained with the entire datasets of
Table 6 through Table 11 that consists of 30 experiment data plus the data from Table 12. Then, three
confirmation tests were conducted to identify the accuracy of the developed ANN model, shown in
Table 13. Hence, to study the effect of joint fit-up disturbances, the final developed ANN model was
trained using 42 experimental data from Table 6 through Table 13.
Table 12: The DOE for the effect of joint fit-up disturbances.
L9 P.D. (mm) Gap (mm) mismatch (mm) misalignment (mm) UTS (MPa)
1 6.1 0 0 0 241
2 6.1 0.25 -0.2 -0.5 238
3 6.1 0.5 0.4 0.5 128
4 6.2 0 -0.2 0.5 235
5 6.2 0.25 0.4 0 155
6 6.2 0.5 0 -0.5 206
7 6.3 0 0.4 -0.5 127
8 6.3 0.25 0 0.5 240
9 6.3 0.5 -0.2 0 245
Table 13: Confirming tests for the ANN model including part.
P.D. Gap mismatch misalignment w V UTS ANN-UTS Error

(mm) (mm) (mm) (mm) (rpm) (mm/min) (MPa) (MPa)
1 6.1 0 -0.08 0 3500 540 240 252 5%
2 6.2 0.25 0.43 0 3500 540 149 157 5.4%
3 6.3 0.25 -0.08 0.5 3500 540 242 241 0.4%
3.5.2 ANN modeling

The neural networks (NN) were trained using the experimental data presented in the previous part.
To model the relationship between the inputs and outputs, feed-forward neural networks with back
propagation algorithm were used. Prediction capability of the back propagation NN model depends
on some factors like the architecture of the network, learning rate, and momentum coefficient [73].
In this paper, the architecture of the networks uses one hidden layer and the number of neurons is
based on estimation errors shown in Table 14. To avoid data overfitting, the number of neurons in
the hidden layer was minimized while keeping the modeling error low. Different architectures,
learning rates, and momentum coefficients were used. After training the root-mean squared error,
RMSE, the mean relative error, MRE, the correlation coefficient, R2, the mean absolute error, MAE,
57
and the maximum error are estimated as indicated in Table 15 and Table 16. Based on the mentioned
criteria, the network architecture to predict UTS, force, and UTS in presence of joint fit-up
disturbances are 3-5-1, 3-6-1, and 6-5-1, respectively. Figure 46 depicts the topology of ANN models.
Moreover, it was found that an adequate setting of learning rate and momentum were 0.5 for both of
them (see Table 14). Table 15 demonstrates the amount of mentioned errors for the final developed
ANN models. It is evident that the trained models has low amounts of error. Another factor to show
the performance of the ANN models is correlation coefficient, R2. It indicates the relationship
between the experimental data and the predicted value by the developed ANN model. Therefore, the
R2 values for the trained ANN models with the optimized training features, as shown in Table 14, are
studied here to evaluate the prediction capability of the models. Table 16 presents the R2 values for
the developed ANN models before the confirmation tests, R2 for confirmation tests, and R2 of the
final ANN models trained using the whole datasets. The R2 values for UTS at training, confirming,
and finalized training are 0.999967, 0.998883, and 0.999968, respectively showing the good
agreement with the experimental results. From Table 16, it is also obvious that the R2 values for force,
and UTS of the joints in presence of joint fit-up disturbances are near unity. The errors of Table 15
and R2 values of Table 16 confirm the accuracy of the final ANN models even in presence of part fit-
up disturbances.
Table 14: Training parameters used for developed ANN models.
Model
Parameter UTS model Force model UTS model for joint fit-up
Network configuration 3-5-1 3-6-1 6-5-1
Number of inputs 3 3 6
Number of hidden layers 1 1 1
Number of neurons 5 6 5
Number of outputs 1 1 1
Total No. of experimental data 30 30 42
Learning rate 0.5
Momentum 0.5
Method Back propagation algorithm
58
(a) (b)
(c)
Figure 46 : The architecture of the developed neural networks (NNs): (a) a 3-5-1 NN for UTS, (b) a 3-6-1 for
downward axial force, (c) a 6-5-1 for UTS in presence of joint fit-up disturbances.
Table 15: Error values for the final ANN models.
R2 R2 R2
∑𝑵
𝒊 (𝑨𝒊 − 𝒀𝒊 )
𝟐
(trained model before final (confirming (final model obtained by all
𝑹𝟐 = 𝟏 − ( )
∑𝑵𝒊 (𝒀 𝒊 ) 𝟐 confirmation) tests) experiments)
UTS model 0.999967 0.998883 0.999968

Force model 0.999532 0.998030 0.999539
UTS model for joint fit- 0.999929 0.998571 0.999786
up
(Ai, Yi, N, and i are experimental value, predicted value, total number of experimental data, and trial number, respectively.
The formula is from Ref.[76]).
Table 16: Correlation of the final ANN models.
RSME MAE MRE Maximum error

1⁄ 𝑁 𝑁
formula 𝑁 2 1 1 |𝐴𝑖 − 𝑌𝑖 |
1 ∑|𝐴𝑖 − 𝑌𝑖 | ∑( ) × 100 |𝐴𝑖 − 𝑌𝑖 |
( ∑(𝐴𝑖 − 𝑌𝑖 )2 ) 𝑁 𝑁 𝐴𝑖
𝑁 𝑖 𝑖
𝑖
UTS model 1.3 MPa 0.9 MPa 0.4 % 3.3 MPa
Force model 0.12 kN 0.09 kN 1.8 % 0.3 kN
UTS model for joint fit-up 3.1 MPa 2.8 MPa 1.4 % 5.4 MPa
The formulas are from Ref.[76]).
59
3.6 Results and discussions
3.6.1 Effect of tool plunge depth
During the FSW process, most of the generated heat is provided by the friction in the
shoulder/workpiece interface [79]. Indeed, sufficient frictional heat must be generated during the
process to accomplish a sound, defect-free weld. An adequate amount of heat will appear when the
amount of generated axial force during the FSW process is sufficiently high [80]. In this regard, effect
of tool plunge depth on the generated force and tensile properties of the joints is studied here. The
experimental results for the generated force and the UTS of RAFSW joints are depicted in Figure 47.
The joints are welded employing optimal tool rotational and traverse speed of 3500 rpm and 540
mm/min, respectively. From Figure 47, it seems when the tool plunge depth reaches to 6.15 mm,
the amount of generated force is enough to generated sufficient amount of heat in the welded area. It
is clear when the tool plunge depth is less than 6.05 mm the UTS is unacceptable. From 6.05 mm to
6.15 mm, the more plunge depth, the higher UTS and force. At the plunge depth of 6.15 mm, the UTS
reaches a maximum of 253 MPa with a ductility of 11.5% while the generated force is as low as 5.9
kN.
(a) (b)
(c)
Figure 47 : The effect of plunge depth on generated axial force and tensile properties of the welded samples at
optimized traverse speed of 540 mm/min, rotational speed of 3500 rpm, and various plunge depths from 5.95
mm to 6.2 mm.
60
3.6.2 Axial force during RAFSW
Figure 48 presents the correlation between the tool plunge depth, the rotational speed, the traverse
speed and the axial force during the RAFSW process as estimated by the final ANN model trained
with all datasets. As can be seen, higher axial forces appear typically with larger plunge depths and
faster rotational speeds. More specifically, for a traverse speed of 540 mm/min, along the horizontal
line in Figure 48.a, the force increases with the penetration depth, which is typical, and decreases over
6.2 mm/min because the metal tends to flow out and generates burrs. Still in the same graph, at the
penetration depth of 6.15 mm, it can be observed when the traverse speed increases, the force
increases at first and then decreases. At low traverse speeds, the heat is more intense cause to
softening of the aluminum. At higher traverse speeds, it seems that the material flows differently
especially as it returns in the joint after stirring which indicates lower pressures under the tool
shoulder and consequently lower axial forces. [1]. Figure 48.b shows, for a given penetration depth,
an increase of force with rotation speed. As the tool rotational speed is increased, the process produces
more heat and the metal flows more easily thus increasing the pressure under the tool shoulder. This
seems confirmed by thickening and the presence of trace marks appearing on the surface of the joint.
(a) (b)
(c)
Figure 48: Contour plots to show the correlation of the generated axial force during RAFSW process with: (a)
tool plunge depth and traverse speed (at constant rotational speed of 3500 rpm), (b) tool plunge depth and
rotational speed (at constant traverse speed of 540 mm/min), (c) tool rotational and transverse speed (at
constant tool plunge depth of 6.15mm). The contour plots extracted from final ANN model for force.
61
3.6.3 Ultimate tensile strength of RAFSW specimens
Figure 49 presents the relationships between the process parameters and UTS of the welded samples
after 7 days of natural aging. Figure 49.a and b shows that good UTS requires a penetration depth of
at least 6.1 mm. By looking at UTS for a penetration depth of 6.15 mm (white vertical line) in Figure
49.a, it can be observed that there is an increase of UTS followed by a drastic drop at 610 mm/min.
Indeed, at low traverse speed, the heat is more intense which softens the material and probably causes
a wider heat affected zone. Indeed, at lower traverse speeds, the heat is more intense which softens
the material and probably cause to a wider heat affected zone around the joint [1]. At speeds exceeding
610 mm/min, the force generated by the tool rotation and heat is insufficient and does not have time
to spread entirely before stirring. Figure 49.b and 6.c provide the rotation speeds needed for good
UTS. This stands between 3150 and 3700 rpm.
(a) (b)
(c)
Figure 49: Contour plots to show the correlation of the UTS of RAFSW joints with: (a) tool plunge depth and
traverse speed (at constant rotational speed of 3500 rpm), (b) tool plunge depth and rotational speed (at
constant traverse speed of 540 mm/min), (c) tool rotational and transverse speed (at constant tool plunge
depth of 6.15mm). The contour plots extracted from final ANN model for UTS.
Globally, it seems that UTS ranging between 231 and 247 MPa (orange regions), are slightly weaker
than the red region (which UTS is more than 253 MPa) due to the excess of heat during welding.
62
Excessive heating appears at high rotational speeds and at low traverse speed. The noticeable regions
in these graphs are where the UTS is 247 MPa and over. If these zones of highest UTS be combined
with the indicated zones of axial force of Figure 47 specially where the force is lower than 6.5 kN,
the working window of parameters will be achieved as shown in Table 17. Within this window, the
force is lower than 6.5 kN while the UTS of the welds is more than 247 MPa. This presented range
for the generated force during RAFSW is significantly lower than the generated forces during FSW
at tilt angle for AA6061-T6 samples [46].
Table 17: Efficient working window of the process parameters of the RAFSW technique
Process parameter Tool plunge depth Tool transverse speed Tool rotation speed
Range 6.125 to 6.3mm 500 to 640mm/min 3220 to 3640rpm
3.6.4 Robustness of the RAFSW process

Industrial use of RAFSW techniques requires robustness meaning that the welding process must
present repeatedly the same good strength despite typical process disturbances [46]. In a previous
research, the capability of compensating joint fit-up disturbances by changing the process parameters
of RAFSW were studied at low traverse speeds [28]. In this section, the tolerable amount of each
joint fit-up variation will be established using the optimized conditions identified previously for
disturbance-free welded samples. Regarding this issue, an ANN model was trained with the data of e
the force is lower than 6 kN.
Table 6 through Table 9, Figure 46, Table 10 and Table 11 to identify the tolerable range of mentioned
disturbances during the FSW at high traverse speeds. Figure 49 presents the UTS of welded samples
at rotational and traverse speed of 3500 rpm and 540 mm/min respectively. Figure 50.a shows the
impact of gap, between the aluminum extrusions, on UTS of the joints. It also shows how the
penetration depth can compensate the lack of material in the gap up to about 0.5 mm. Without
compensation and using 6.15 mm of penetration depth, only a small gap of 0.1 mm can be tolerated.
As an instance, when the plunge depth is 6.22 mm, the process could compensate the negative effect
of a 0.5mm gap. In this connection, Cole at al. and Shultz et al. [22, 46] reported similar results with
regular FSW. Another variation in the process is the mismatch. When the advancing side is higher
than the retreating side, the mismatch is positive, and vice-versa. Based on previous research [28],
positive mismatch is more acceptable than negative. Thus, this fact is considered it in the design of
experiments in this research. From Figure 50.b, it is clear that at the plunge depth of 6.15 mm, from
negative mismatch of -0.2 mm through a positive mismatch of 0.25 mm, the UTS would be still in an
acceptable range. Besides, the more the tool plunge depth, the wider the range of tolerable mismatch.
63
Figure 50.c shows the correlation between misalignment, tool plunge depth, and UTS. As can be
observed, more misalignment in the advancing side could be tolerated by the process than the
retreating side. Such asymmetric behavior of misalignment compensation was reported also in some
previous research [34, 81, 82]. To illustrate the concept, take for example a plunge depth of 6.15 mm.
As can be seen, the process could tolerate -0.4 mm misalignment in the advancing side, while it could
tolerate just 0.2 mm of misalignment in the retreating side. Furthermore, using higher plunge depths
the process can tolerate wider range of misalignments. Figure 50.d shows the relationship between
the simulated thickness variation for the workpiece material, tool plunge depth, and UTS. It is evident
that positive variations are more desirable than negative ones, especially with lower plunge depths.
Additionally, the more plunge depth, the more negative variation of thickness would be tolerable.
(a) (b)
(c) (d)
Figure 50: Contour plots to show the correlation of the UTS of RAFSW samples with: (a) gap and tool plunge
depth, (b) mismatch and tool plunge depth, (c) misalignment and tool plunge depth, and (d) thickness
variations of the plates and tool plunge depth. All RAFSW samples are welded at optimized tool rotational
speed of 3500 rpm and traverse speed of 540 mm/min. The contour plots extracted from final ANN model for
UTS of welded samples in presence of joint fit-up disturbances.
Generally, it shows that with deeper plunging, wider range of disturbances are tolerable. However,
utilizing deeper tool plunging leads to higher axial forces, large burrs and excessive flashes.
64
Consequently, it is important to select the plunge depth efficiently or to control it. Figure 51 shows
the UTS, as two types of variations are present in the welding setup. FSW was conducted at 6.15 mm
of plunge depth, 3500 rpm, and 540 mm/min transverse speed. As expected, it shows that smaller
variations are tolerable when more than one type of variations is present at the same time.
(a) (b)
(c)
Figure 51: Contour plots to show the correlation of the UTS of welded samples with two types of disturbances
at the same time which are: (a) mismatch and gap, (b) misalignment and gap, and (c) misalignment and
mismatch. All RAFSW samples are welded at optimized tool rotational speed of 3500 rpm and traverse speed
of 540 mm/min. The applied tool plunge depth was 6.2 mm. The contour plots extracted from final ANN
model for UTS of welded samples in presence of joint fit-up disturbances.
3.6.5 Effect of artificial aging

Post weld heat treatment (PWHT) is an attempt to recover some portion of lost mechanical properties
of welds [1]. Using samples welded under optimal conditions (PD= 6.15 mm, v= 540 mm/min, w=
3500 rpm) and characterized by a UTS of 253 MPa and 11.5% ductility after natural aging, the effect
of post held heat treatment is studied. The PWHT results reveal that the artificial aging process can
significantly boost the UTS and yield strength of FSW samples up to 300 MPa and 260 MPa,
respectively. However, it has an adverse effect on the ductility that dropped to 4.3%.
65
3.6.6 Macrostructure examinations
Figure 52 presents a sample welded in optimal conditions. We can see that the surface of the joint
has no burrs and is aligned to the surface of the sample. Moreover, there is no depression in the surface
of the joint. This is not true with FSW joint using a tilt angle. In Figure 52.b a cut section hand
polished and macro-etched shows the welded nugget. Indeed, the depression of the surface in the
welded area reduce the strength of the joint as mechanically loaded. Additionally, the sharp radius at
the edge of the joint is the origin of stress concentration especially important under cyclic loading.
Hence, the absence of sharp edges and depression on the surface of the RAFSW joints could be
considered as a positive aspect of this technique.
(a)
(b)
Top line
(c) Centerline
Bottom line
Figure 52: (a) Top surface view of the RAFSW joint welded at optimized condition consists of tool plunge
depth of 6.15 mm, rotational speed of 3500 rpm, and traverse speed of 540 mm/min, (b) macro structure of
the RAFSW joint which shows the nugget zone shape with unarmed eyes, (c) schematic of FSW joint
illustrates distinct microstructural regions called nugget zone (NZ), Thermomechanically affected zone
(TMAZ), heat affected zone (HAZ), and base material (BM).
3.6.7 Microstructure examinations

The microstructure of the base metal, the heat-affected zone (HAZ), the thermo-mechanically-
affected zone (TMAZ) and the nugget zone (NZ) appears in Figure 53. It illustrates a sample welded
using optimal conditions as indicated previously. Figure 53.a shows the elongated grain structure of
the base metal. The HAZ microstructure is similar to base material with slightly coarser grains, as
illustrated in Figure 53.b. The small area of the TMAZ zone shows a distorted microstructure (see
Figure 53.c). Finally, the nugget zone in the weld reveals an equiaxed grain structure, as can be seen
in Figure 53.d. This could be attributed to the dynamic recrystallization during the FSW process [1].
66
(a) (b)
(c) (d)
Figure 53: Microstructure of (a) base metal, (b) HAZ, (c) TMAZ, and (d) NZ, at 500X magnification. The
RAFSW joint welded at optimized condition consists of tool plunge depth of 6.15 mm, rotational speed of
3500 rpm, and traverse speed of 540 mm/min.
3.6.8 Micro-hardness examinations

Figure 54 illustrates the hardness profile across the welded joint. The measurements were taken at 1
mm intervals along horizontal lines traversed at the top, center, and bottom of the metallographic
weld specimen, as shown in Figure 52.c. The obtained micro-hardness profile enclosed the heat-
affected zone, HAZ, thermomechanical affected zone, TMAZ, and nugget zone, NZ. The hardness
profile is strongly affected by precipitates size and their distribution [1]. As can be seen in Figure 54,
the hardness profile has a W shape, which is an identical characteristic of FSW joints. The lowest
hardness in top, center, and bottom lines are around 5.5, 5, and 4 mm away from the vertical center of
the weld, respectively. The nugget zone and TMAZ are located between the two minimum points in
the hardness profile. During FSW process, the NZ region undergoes intense plastic deformation at
high temperatures near the melting point of material. This causes to dynamic recrystallization of the
grains and dissolution or coarsening of precipitates [37] Thus, it has a fine equi-axed microstructure
as can be seen in Figure 53.d. Indeed, in spite of dissolving or coarsening of precipitates in this region,
it has a higher hardness than TMAZ zone, as can be seen in Figure 54, due to smaller grain size and
higher density of sub-grains [12]. In TMAZ, the material undergoes plastic deformation, Figure 53.c.
The deformed grains in this area contain a high density of grain boundaries [1]; and most of
precipitates are dissolved [1, 37]. HAZ region has nearly the same microstructure as the base material,
as shown in Figure 53. The main reason for lower hardness of HAZ than base material is regarded to
the dissolving or coarsening of precipitates in this area [1, 37]. The tensile test sample fractured in
the minimum hardness location that was in the retreating side of the sample. Based on the
67
metallographic observations, the lowest hardness is located somewhere between the TMAZ and HAZ
zones. Moreover, the effect of PWHT on the hardness profile of the examined sample is depicted in
Figure 54. It reveals that artificial aging leads to significant improvement of hardness of the joint in
HAZ, TMAZ, and NZ regions through re-precipitation mechanism.
Figure 54: Hardness profile for the optimal sample along the top, center, and bottom lines as indicated in
Figure 52. Besides, the hardness profile of the PWHT sample along its centerline. The RAFSW joint welded
at optimized condition consists of tool plunge depth of 6.15 mm, rotational speed of 3500 rpm, and traverse
speed of 540 mm/min.
3.7 Conclusion
Based on the results obtained in this research, the main conclusions can be summarized as follows:
• The effect of process parameters on generated force and UTS of the joint has been studied.
Additionally, the best working window providing maximum UTS with low forces on butt-
welding of AA6061-T6 has been established for RAFSW. In the provided working window,
welding of 6.35 mm thick bars can be done at a traverse speed as high as 600 mm/min to reach
UTS of more than 247 MPa.
• The robustness of the process is demonstrated by testing the effect of different geometric errors,
without any modification of the pre-determined process parameters. Defect-free welds with
acceptable tensile strengths can be achieved, repeatedly.
• The post weld artificial aging at 160°C for 18 h could recover a great portion of lost tensile
properties of the weld. The UTS of the artificially aged sample boosts from 253MPa to 300 MPa.
68
• Macro-structural investigations indicates the absence of burrs at the edges and no surface
depression around the weld area that could positively affect the mechanical properties of the
RAFSW joints under the static and dynamic loadings.
• Finally, the microstructure of the welds and the profiles of Vicker’s micro-hardness for the
optimized joint has studied. The results validate that the optimized sample is defect-free and has
a normal w-shape micro-hardness profile.
The results provided in this research is a full guideline regarding the characteristics of the recently
introduced RAFSW technique applicable industrially on common inexpensive 3-axis CNC machines.
69
Post Weld Heat Treatment effects on
Mechanical Properties and Microstructure of
AA6061-T6 Butt Joints Made by Friction Stir Welding
at Right Angle (RAFSW)

Keywords: Friction stir welding at right angle; Post weld heat treatment; Aging; Solubilizing; Tensile
properties; Hardness
This Chapter has been published in: Journal of Manufacturing and Materials Processing, Vol 3 (2),
42 (2019)
https://doi.org/10.3390/jmmp3020042
Licensee MDPI, Basel, Switzerland. This is an open access article distributed under the Creative
Commons Attribution License which permits unrestricted use, distribution, and reproduction in any
medium, provided the original work is properly cited.
(https://creativecommons.org/licenses/by/4.0/)
70
4.1 Résumé
Le soudage par friction-malaxage (FSW) offre aux utilisateurs de nombreux avantages par rapport
aux techniques de soudage par fusion. Néanmoins, il n'est pas largement utilisé dans l'industrie
actuelle, principalement en raison des coûts d'équipement élevés et des redevances. Pour surmonter
ces problèmes, une technique FSW à faible coût opérée à angle droit, appelée RAFSW, a été
développée récemment par notre équipe de recherche. Pour rendre la technique RAFSW fiable pour
les utilisateurs potentiels, nous allons analyser l'effet de divers traitements thermiques post-soudure
(PWHT) sur les propriétés mécaniques et physiques des joints RAFSW. À cette fin, des paramètres
de processus optimisés sont utilisés pour souder des joints bout à bout en alliage AA6061-T6. Les
joints ont été caractérisés en utilisant un essai de traction, un essai de micro-dureté et des techniques
de métallographie. Le temps de vieillissement le plus efficace a été obtenu pour différentes
températures de vieillissement. De plus, il a été constaté que le vieillissement artificiel à 220 ° C
pendant 30 min pouvait être utilisé comme un PWHT de vieillissement artificiel rapide et rentable
pour le secteur industriel. De plus, la répétabilité des PWHT a été démontrée en étudiant l'effet du
temps d'attente avant le vieillissement artificiel. Enfin, il a été révélé qu'un seul processus de
vieillissement artificiel rapide est plus avantageux que la solubilisation suivi d'un processus de
vieillissement artificiel en termes de propriétés de traction, de temps consommé et de coût.
4.2 Abstract
Friction stir welding (FSW) provides users with many advantages over fusion welding techniques.
Nevertheless, it does not widely employed in current industry mainly due to high equipment costs
and royalties. To overcome these issues, a low-cost FSW technique operated at right angle, called
RAFSW, has been developed by our research team, recently. To make the RAFSW technique reliable
for potential users, we are going to analyze the effect of various post weld heat treatments (PWHT)
on mechanical and physical properties of the RAFSW joints. To this end, optimized process
parameters are used to weld butt joints of AA6061-T6 alloy. The joints were characterized using
tensile test, micro-hardness test, and metallography techniques. The most efficient aging time was
obtained for various aging temperatures. Moreover, it was found that artificial aging at 220°C for 30
min could be used as a fast and cost-effective artificial aging PWHT for industrial sector. In addition,
the repeatability of the PWHTs were demonstrated by studying the effect of waiting time prior to the
artificial aging. Finally, it was revealed that a single fast artificial aging process is more beneficial
than solubilizing followed by artificial aging process in terms of tensile properties, consumed time,
and cost.
71
4.3 Introduction
Friction stir welding (FSW) has attracted an increasing interest among researchers and industrial
sectors over the past two decades [1]. Compared to the common fusion-welding techniques, like MIG
welding, FSW provides users with many advantages such as better mechanical and fatigue properties
of the weld, ability to weld dissimilar alloys and non-weldable alloys by fusion welding techniques,
less energy-consumption, and environmental friendliness [1, 2, 4, 8, 19, 83, 84]. However, there are
some obstacles to its prevalent use such as the need for CNC equipment, the large forces involved,
the need to use sturdy fixtures, and generally the equipment costs and the royalties [8, 18, 19]. To
have a cost-effective FSW process, PI2/REGAL team at Laval University has recently developed a
low-cost FSW technique operated at right angle [20, 24, 27-29, 70]. In this technique, called RAFSW,
common low-cost 3-axis CNC machines, which are used for machining, are utilized for RAFSW
technique too, without any modification of the CNC machines.
Another barrier against the widespread use of FSW in industry is the fact that there are a lot of
uncertainties, scientific gaps, and lack of guidelines regarding other aspects of this process like post
weld heat treatment (PWHT) of FSW joints [19]. Although it is confirmed that PWHT could greatly
enhance the mechanical properties of the FSW joints [48-50], the reported research is limited. Indeed,
there is lack of research on the effect of PWHT on the FSW joints, especially for welded samples by
FSW at right angle and at high traverse speeds. Therefore, it is essential to shed light on the impact
of various PWHTs processes on the optimized welded samples by RAFSW technique to make it
reliable for potential industrial users. Various parameters affect the physical and mechanical
properties of the FSW joints such as microstructure, tensile properties, and micro-hardness
distribution of the weld area [83, 85, 86]. Tool design is one of the influential parameters since the
mixing and stirring mechanisms within the weld area are in control of that [87]. Moreover, the
generated forces during the process and the process parameters including tool tilt angle, plunge depth,
rotational speed, and traverse speed affect the physical and mechanical properties of the joint. Because
they control mixing and stirring process, amount of generated heat, and cooling rate during the FSW
process [1, 83]. The developed RAFSW technique has its own features in terms of tool design,
involved forces, and optimized process parameters. Therefore, the physical and mechanical properties
of the welds are affected by all of them [70]. Thus, it is necessary to investigate and characterize the
effect of different natural aging, artificial aging, and solubilizing processes on the RAFSW joints to
make the process reliable for industrial use.
Therefore, the main objective of the present paper is to study the effect of various PWHTs on
mechanical and physical properties of the RAFSW joints welded under optimized process parameters
72
at high traverse speeds obtained in our previous research [70]. To make the welds, we have utilized a
recently developed RAFSW technique using low-cost 3-axis CNC machine tools [20, 24, 27-29, 70].
4.4 Materials and Methods

4.4.1 Preparation of the joints
In this research, the RAFSW technique is employed to butt weld extruded AA6061-T6 flat bars, 6.35
mm thick, 76.2 mm wide, and 250 mm long cut by a power hacksaw. Based on the extruder
specification for the as-received AA6061-T6 alloy, it has a nominal chemical composition of 0.83
wt.% Mg, 0.55 wt.% Si, 0.19 wt.% Cu, 0.19 wt.% Fe, 0.05 wt.% Cr, 0.07 wt.% Mn and the rest is
aluminum. Tensile test results of the as-received material shows that the ultimate tensile strength and
ductility are 285 MPa and 16.4%, respectively. To make the joints, two flat bars are fixed onto a rigid
back-plate. The back-plate is placed on a 3-axis Kistler 9265B dynamometer fastened to the table of
a 3-axis CNC machining center. The utilized 3-axis CNC machine is a Fryer MC-15 which has a 25
HP spindle using CAT40 tool holders with the maximum rotational speed of 8000 rpm and the axis
peak trust of 15 kN. To conduct RAFSW, a tool designed specifically for this purpose is mounted
into a long reach tool holder. The shoulder diameter of the tool and its pin length are 15 and 6 mm,
respectively. The pin shape is spiral and some grooves are designed on the shoulder. Figure 55.a and
b demonstrate the mentioned installation, the utilized tool, and the tool holder. The described set-up
and installation is exactly what was used in the previous research regarding the optimization of the
RAFSW technique [70]. The obtained optimized process parameters are used to weld the samples
according to our previous research [70]. The tool plunge depth, rotation speed, and traverse speed
were 6.15 mm, 3500 rpm, and 540 mm/min, respectively [70]. Figure 55.c. depicts the top view of
the single-pass welded joint along the extrusion direction of the bars under the mentioned process
parameters.
4.4.2 Post weld heat treatment

A calibrated PYRADIA furnace equipped with a monitoring system for temperature variations is used
for PWHT processes. The targeted PWHTs are different combinations of the natural aging, artificial
aging, and solubilizing processes. The effect of natural aging time on as-weld samples were studied,
firstly. Then, the impact of various artificial aging processes were characterized at 160, 180, 200, 220,
and 240 °C for various periods from 20 min to 18 h on the as-weld samples. Moreover, the effect of
the natural aging prior to the artificial aging was studied to evaluate the repeatability of the PWHT
results when there is a delay between welding and artificial aging process. Finally, the effect of the
solubilizing was studied at 530 °C for 1 h followed by the optimized artificial aging called T6,
73
obtained from the previous set of experiments. For each PWHT process, two samples, one welded
sample and one raw material sample, were subjected to the designed heat treatments.
4.4.3 Characterization
Tensile test was done to evaluate the tensile properties of the joints. The uniaxial tensile test was done
according to the standard of American Society for Testing of Materials, ASTM E8M-04 standard.
The geometry of the machined samples is illustrated in Figure 56. For that, a hydraulic testing
machine with a 44.5 kN load cell was used. The crosshead speed of the testing machine was 1
mm/min. Moreover, a high-resolution Epsilon extensometer was used during the tests. For
metallographic investigations, welds are cut around the joint area and mounted in the phenolic resin.
Then the mounted samples are polished to 2000 grits SiC papers followed by final polishing by
colloidal silica suspension using an automatic Buehler polishing machine. Afterwards, the polished
samples were pre-etched with 10 wt% NaOH solution followed by etching with Weck’s reagent [70,
78]. Then a Nikon K100 optical microscope was used to take the micrographs. The macrograph of
the joint was captured using a High-resolution camera. The used macro-etchant was a solution
containing 75 ml HCl+ 25 ml HNO3+ 5 ml HF diluted with 25% distilled water. Consequently, a
Buehler Vickers micro hardness-testing machine was used to measure the micro hardness of the
samples within and around the joint. The measurements are done according to the ASTM E384
standard by applying a 300 g load and a dwell time of 15 s.
(a) (b)
FSW tool Clamps
Abutted edge
Aluminum bars
Back-plate Dynamo-meter (c)
Figure 55: (a) The set-up of welding consist of the extruded bars, the clamping system and the Kistler 3-axis
dynamometer; (b) FSW tool and tool holder; (c) top view of the welded sample by optimized process
parameters [70].
74
Figure 56: Schematic of tensile test samples.
4.5 Results and Discussion

4.5.1 Natural aging
Figure 57 depicts the effect of natural aging time on the tensile properties of the as-welded RAFSW
samples. It indicates that the tensile properties drastically change until 7 days of natural aging.
Moreover, it can be inferred that the tensile properties are almost stable after 14 days of natural aging.
The tensile strength increases by passage of time while the ductility decrease. This is attributed to the
kinetic of the precipitation in different regions of the joint. Precipitates evolution leads to partial
recovery of strengthening through the welded area [88]. As a results, the tensile strength of RAFSW
joint boosts in the course of time. Generally, the stronger joint, the less ductility it will have, which
is confirmed with our results as shown in Figure 57. The stabilized tensile strength, ductility, and
joint efficiency of the as-weld sample are 204 MPa, 14.2 %, and
72 %, respectively.
(a) (b
Figure 57: The effect of natural aging time on tensile strength: (a) and elongation; (b) of welded samples at
optimized process parameters by RAFSW.
Figure 58 demonstrates the macrostructure and micro-hardness map of the naturally aged RAFSW
sample. Distinct regions of FSW welds called nugget zone (NZ), thermo-mechanically affected zone
(TMAZ), heat affected zone (HAZ), and unaffected base material (BM) are identified in Figure 58.a.
The micro-hardness map across the weld area, shown in Figure 58.b, is in conjunction with the
identified zones in the macrostructure shown in Figure 58.a. Generally, the hardness drops in the weld
area for the FSW joints of AA6061-T6, as can be seen in Figure 58.b. This behavior is mainly
associated with the dissolution or coarsening of the precipitates within the weld area during the
75
welding process and re-precipitation and coarsening of the precipitates during the subsequent aging
process [88]. Figure 59 demonstrates the microstructure around the NZ. The borders of the NZ,
TMAZ, and HAZ is obvious in the advancing side while is not distinguishable in the retreating side
of the RAFSW samples. This asymmetric behavior originates from the asymmetric effect of the
rotating tool inside the material [89].
(a)
(a) BM HAZ TMAZ NZ TMAZ HAZ BM
(b)
Figure 58: (a) Macrostructure at the cross-section of the joint marked with the distinct weld regions; (b)
Micro-Vickers hardness map around the welded region of a naturally aged sample made by RAFSW
technique operated at optimized process parameters.
(a) (b)
HAZ TMAZ NZ
Figure 59: Microstructure of welded sample around the nugget zone at: (a) advancing side; (b) and retreating
side for naturally aged joints made by RAFSW at optimized process parameters.
As can be seen in Figure 60.a to c, the NZ shows a fine, equi-axed microstructure compare to the BM
because of dynamic re-crystallization and formation of special grain boundaries during the FSW
process [90]. The microstructure of the NZ changes from top to the bottom of the nugget zone. The
finest grains are observed in the top and the biggest in the middle near the bottom, Figure 60.a to c.
This could be attributed to both mixing and stirring mechanisms in different areas of the NZ and the
difference in the amount of heat input which is more in top-side due to the contact with the tool
shoulder. Moreover, the difference in cooling rate plays a role in the final microstructure of the grains
and the precipitates evolution during and after the welding process [91]. The finer microstructure, the
higher micro-hardness. In this connection, the microstructure observations for the NZ is in
conjunction with the micro-hardness profile shown in Figure 58.b. From Figure 60.d to f it is obvious
76
that the grain structure in the HAZ area far from the NZ seems like BM, while in the HAZ near the
NZ, the grains are relatively smaller and less elongated than the HAZ far from the NZ. It could be
originated from the difference in experienced thermal cycle during the weld. The micro-hardness
map, Figure 58.b, demonstrates the HAZ regions far from the NZ has much higher hardness than the
HAZ regions near the NZ area. It could be assigned to the major difference in precipitates
morphology, size and distribution in these areas.
(a) (b) (c)
(d) (e) (f)
Figure 60: Microstructure of: (a) top; (b) middle; (c) and bottom of NZ zone; (d) base material; (e) HAZ area
far from weld region; (f) and HAZ area near NZ for naturally aged joints made by RAFSW.
4.5.2 Artificial aging

There is lack of study on the effect of artificial aging at different time and temperatures for FSW
samples especially at zero-tilt angle and high traverse speeds. While, the industry is in need of
reliable, cost-effective PWHTs. Therefore, in this part of research, different artificial aging processes
on RAFSW butt-joints are examined. In this regard, Figure 61 demonstrates the effect of artificial
aging time and temperature on the tensile strength of the RAFSW welded samples. Generally, the
higher aging temperature, the lower time is needed for precipitates evolution [92]. It is obvious that
at first the tensile strength of the samples increases with aging time at each aging temperature. This
originates from strengthening mechanism of evolved precipitates in the weld area due to the aging
process. As can be seen in Figure 61, after reaching a peak of tensile strength at a certain aging time
for a given aging temperature, the tensile strength starts to decrease due to over aging [92]. As it is
observable, this reduction is sharper for higher aging temperatures due to the fast kinetics at higher
temperatures. It can be concluded that the highest joint efficiency and tensile strength under aging at
180°C for 18 h, are 90 % and 257 MPa, respectively. The results of the artificial aging at 160 °C for
18 h was approximately similar. The tensile strength in this case was as high as 254 MPa. Since the
77
industry is in need of fast and cost-effective PWHTs, one can conclude that a fast, cost-effective
artificial aging process at 220 °C for 30 min is quite beneficial. In this case, the joint efficiency, and
tensile strength of the RAFSW joints would be as high as 85 % and 241 MPa, respectively. Table 18
provides a summary on the tensile properties of the plain material, naturally aged weld, and artificially
aged welds at optimized conditions obtained from Figure 61. In macro-scale, the elongation of the
samples is shown in Table 18.
(a) (b)
(c) (d)
Figure 61: The effect of artificial aging at: (a) 180°C; (b) 200°C; (c) 220°C; (d) and 240°C under different
aging times on the tensile strength of the RAFSW samples.
Table 18: Tensile properties and joint efficiency of the welds aged naturally and at optimized artificial aging
conditions.
Samples condition UTS (MPa) El (%) Joint efficiency 2
Plain Material (as received) 285 16.4 -

As-weld 188 15.9 66 %
NA1 for 14 days 204 14.2 72 %
AA1 at 160°C for 18 h 254 4.1 89 %
AA at 180°C for 18 h 257 4.2 90 %
AA at 200°C for 2 h 245 6 86 %
AA at 220°C for 30 min 241 6.1 85 %
AA at 240°C for 30 min 230 6.6 81 %
1NA and AA are abbreviations for naturally aged and artificially aged, respectively.
2Joint efficiency is the ratio of the strength of the joints to the strength of the plain material in percentage.
78
Figure 62 demonstrates the micro-hardness distribution at centerline through the weld area of
naturally aged joint, artificially aged at 160°C for 18h, and recommended fast aged sample at 220°C
for 0.5 h. Indeed, during the subsequent artificial aging process on welded joints, the dislocations
density and the amount of solute for precipitation differs in different zones of weld. These factors can
mainly control the re-precipitation kinetics. Moreover, the time and temperature of the aging process
affect the precipitation evolution in different zones [93]. As a result, the amount of hardness recovery
in different zones of the weld would be different. Based on these facts, it can be seen that the naturally-
aged sample shows a normal w-shape distribution of hardness as shown in Figure 62. The minimum
hardness is within the HAZ area near the TMAZ. This is in accordance with the minimum recovery
of hardness by precipitates hardening in this area [88]. The NZ area demonstrates a higher hardness
which could be due to its finer microstructure than the HAZ area located near the NZ region.
However, in both area the precipitation hardening mechanism is mostly disappeared due to
dissolution or coarsening of precipitates under thermal cycles of the FSW process [88]. From Figure
62, also it can be concluded that an artificial aging process at 160°C for 18 h leads to recovery of
hardness. It could be assigned to precipitate-strengthening mechanisms by re-precipitation and
precipitate evolutions [50, 94]. In the NZ region, the hardness is considerably restored owing to the
effective type, size, morphology, and homogenous distribution of the precipitates evolved during the
re-precipitation process by aging [95]. This restoration of hardness partially occurred in the TMAZ
and less in the HAZ area near the NZ as shown in Figure 62. It is obvious that artificial aging at 220
°C for 30 min yields less hardness recovery than artificial aging at 160 °C for 18 h. This could be due
to the differences in the number, size, morphology, and distribution of the evolved precipitates.
Because the evolution of the precipitates is a diffusion based mechanism, which is significantly
affected by time and temperature of the aging process [50, 92].
Figure 63 illustrates the microstructure around NZ area of RAFSW joint aged artificially at 160°C
for 18 h. It is evident that the transition between distinct weld zones is more distinguishable in this
sample compare to the naturally aged sample, Figure 59. Microstructural observations indicates there
is no significant difference in the grain structure of the best artificial aged samples at each aging
temperature; including 160 °C for 18 h, 180 °C for 18 h, 200 °C for 2 h, 220 °C for 30 min, and 240
°C for 30 min. They all show a similar pattern under optical microscope, like what we see in Figure
64. Indeed, the difference in their mechanical properties roots in the difference in the type, size,
morphology, and distribution of the precipitates evolved in their different weld zones during the
artificial aging process. In addition, the microstructure of the naturally aged sample shown in Figure
60 looks like artificial aged microstructure of Figure 64, which means no considerable change
happens in the shape and size of grains by the conducted artificial aging processes.
79
Figure 62: Hardness distribution along the centerline of the cross-section of the joints for: (a) naturally aged
RAFSW sample; and for artificially aged samples at: (b) 160°C for 18h and; (c) 220°C for 30 min.
(a)
a (b)
HAZ TMAZ NZ
Figure 63: Microstructure of welded sample around nugget zone at advancing side (a) and retreating side (b)
for RAFSW joint aged artificially at 160 °C for 16 h.
(a) (b) (c)
(d) (e) (f)
Figure 64: Microstructure of top (a), middle (b), and bottom (c) of NZ zone, base material (d), HAZ area far
from weld region (e), and HAZ area near NZ (f) for RAFSW joint aged artificially at 220 °C for 30 min.
80
Based on Table 18 and Figure 62, although artificial aging process boost the strength of the joints and
the hardness in the weld zone, it cannot recover the elongation of the weld samples. It can be attributed
to the differences in the microstructure of the artificially aged joint compare to plain material.
Moreover, it can be assigned to the inconsistency in the microstructure of different zones of the
artificial aged weld samples, as shown in Figure 64. Moreover, the size, morphology, and distribution
of the precipitates are affected by the experienced thermal cycles during and after the welding process.
Precipitates act as barriers against movement of dislocations during the plastic deformation [88-91].
Thus, it is expected that different zones of the welds show different local mechanical properties such
as elongation and plasticity [88, 96]. It is reported that the indentation micro-hardness results
demonstrate a good correlation with local mechanical properties of the weld zones [85, 96].
According to Figure 62, indentation results by micro-hardness diamond probe demonstrates the
difference in the behavior of weld zones. In all samples, the diamond probe had left the biggest
indentation prints in the HAZ area which can indicate that the material in HAZ area demonstrates
more plasticity than other zones. This characteristic of HAZ makes it prone to develop necking and
breakage during the tensile test [85, 86]. This conclusion is in agreement with our experimental results
as all aged joint necked and broke in HAZ area. Moreover, the identified relationship between the
hardness and toughness of the FSW samples can be noted. Indeed, the less hardness of the material,
the higher toughness is expected [97]. Based on Figure 62, it can be concluded that the HAZ area
shows the highest toughness which makes it the most probable area to develop necking and happening
the failure during the tensile test. This idea is also in accordance with our tensile test samples which
all necked and failure in the HAZ area.
In addition, in industrial applications, a delay between the welding process and PWHT of the samples
might happen. During this delay, welded joins are aging naturally. Therefore, evaluating the
repeatability of the PWHTs with different durations of natural aging before artificial aging is
important from industrial perspective. The repeatability of the presented processes in Table 18 is
evaluated in this part. Accordingly, the artificial aging at 160 °C for 18 h, 180 °C for 18 h, 200 °C for
2 h, 220 °C for 30 min, and 240 °C for 30 min was repeatedly done on RAFSW samples after different
natural aging times, from less than one day to 21 days, prior to artificial aging. The results, shown in
Table 19, demonstrate that natural aging prior to artificial aging only slightly affects the tensile
properties of the joints. These results clearly validate the repeatability of the obtained artificial aging
processes when natural aging happens before that.
81
Table 19: Tensile strength of the welded samples subjected to natural aging for less than one day to 21 days
prior to artificial aging process conducted at different conditions based on the obtained results in the previous
part of the study.
UTS (MPa) after artificial aging at different conditions
At 160 At 180 At 200 At 220 At 240

(°C) (°C) (°C) (°C) (°C)
for 18 h for 18 h for 2 h for 30 for 30
min min
Natural aging time (days) prior to 1≤ 254 257 245 241 230
artificial aging 7 251 257 249 243 225
14 249 255 251 238 229
21 250 252 246 238 230
Average UTS and standard deviation 251.0±1.9 255.2±2.0 247.7±2.4 240.0±2.1 228.5±2.1
4.5.3 Solubilizing followed by artificial aging (W+T6)

Another PWHT done in this research is a solubilizing process followed by artificial aging called T6.
In this regard, there is an interesting question among researchers on how a full W+T6 heat treatment
affects a FSW joint; and whether it is possible to recover the lost mechanical properties of the welds
mostly via a T6 heat treatment. Although, a W+T6 heat treatment is costly for most of the industrial
applications, the costs could be justifiable for some applications if the W+T6 can provide better results
than a single artificial aging process.
Accordingly, we have studied the impact of a standard solubilizing process at 530 °C for 1 h followed
by various artificial aging processes at 160 °C for 18 h, 180 °C for 18 h, 200 °C for 2 h, 220 °C for
30 min, and 240°C for 30 min based on optimized conditions of artificial aging presented in Table
18. The findings for tensile strength of the joints are illustrated in Figure 65. It is observable that
artificial aging at either 200 °C for 2 h or 180 °C for 18 h after solubilizing process yield the best
results for RAFSW joints which are solubilized prior to artificial aging, while for plain material
artificial aging at 180 °C for 18 h provides the best results. The results for plain material is in
accordance with the reported data [92]. Figure 65 illustrates that the best condition for a RAFSW
sample is not necessarily the same for the plain material. This sourced from their different
microstructure. A notable point for RAFSW samples is that a relatively fast aging process at 200 °C
for 2 h after solubilizing at 530 °C for 1 h can achieve the best tensile strength among welds that are
solubilized followed by artificial aging. This optimized time and temperature is different from the
82
optimized aging time and temperature when there is no solubilizing process prior to artificial aging.
It could be associated to the fact that solubilizing process changes the driving force for re-precipitation
kinetics and recrystallization of the grains. As a result, there would be differences between the
precipitates’ size, distribution, and morphology at the different weld zones when the welds are
solubilized followed by artificial aging than just artificial aged welds [49, 98-100]. The results also
indicates that at relatively long aging times, the tensile properties of the base metal of the welded
samples starts to degrade, as shown in Figure 65. That could contribute partly to reduction of the
tensile properties of joints due to reduction of the BM properties at aging temperatures higher than
200 °C.
Figure 65: The effect of solubilizing heat treatment followed by artificial aging at different temperatures (at
optimized durations obtained from the previous section) on the tensile properties of the welded samples and
plain material.
Microscopic observations demonstrates that coarsening and abnormal grain growth (AGG) has
occurred within the NZ area in all samples. Figure 66 depicts the microstructure around the NZ,
TMAZ, and HAZ areas of the RAFSW sample solubilized and artificially aged at 200 °C for 2 h after
welding. The size of grains in TMAZ and HAZ areas do not change considerably, while the NZ area
depicts AGG. In one hand, FSW process makes the grains in NZ very fine compare to BM because
of dynamic re-crystallization mechanisms. This causes to a thermodynamic instability inside the
material and encourages the material to stabilize through grain growth. On the other hand, the pinning
force by precipitates against grains growth reduces during the solubilizing process after welding
because of dissolution of precipitates during solubilizing heat treatment of FSW samples [49, 98-
100]. As a result, AGG occurs in NZ region during T6 process, as can be seen in Figure 66. The level
of coarsening is considerably affected by the amount of thermodynamic driving force, pinning force,
83
the microstructure, the time and temperature of the solubilizing process, and the exact chemical
composition [99]. The best W+T6 process is obtained as a solubilizing at 530 °C for 1 h followed by
an artificial aging at 200 °C for 2 h. The joint efficiency and tensile strength for this RAFSW joint
are 85 % and 243 MPa, respectively.
(a) (b)
HAZ TMAZ NZ
Figure 66: Microstructure around (a) and within (b) the nugget zone of the welded sample after W+T6 heat
treatment.
Although applying the obtained W+T6 process results in recovery of the tensile strength of the
RAFSW joint by a notable percentage, the joints have shown low ductility compared to just artificially
aged samples. Low ductility in these samples can be attributed to the formation of precipitate free
zones (PFZs) and AGG within the NZ area during T6 heat treatment [49, 99, 100]. All of the RAFSW
samples in this research fractured within the NZ during the tensile test. That could be attributed to the
fact that the fracture of these samples mostly initiates from the PFZs inside the NZ zone [49, 98].
Therefore, it seems that a full W+T6 heat treatment enhances the tensile strength of the joints at the
expense of their lower ductility. In summary, for RAFSW joints, conducting a solubilizing heat
treatment followed by artificial aging does not yield better mechanical and physical properties than
just an artificial aging process. Moreover, a solubilizing process can cause some distortions and
dimensional errors especially for complex shapes which require machining and finishing processes
after heat treatment. Therefore, a single artificial aging at 220 °C for 30 min on RAFSW samples can
yields not only nearly the same tensile strength, but also the higher ductility than an optimized W+T6
heat treatment consisting of 1 h of solubilizing at 530 °C followed by artificial aging at 200 °C for 2
h. Moreover, that single artificial aging process on RAFSW samples is more time and cost-effective
and presents lower thermal distortion rather than a full W+T6 heat treatment.
84
4.6 Conclusions
• A comprehensive study was done on the effect of natural aging, artificial aging, and T6 heat
treatments on the mechanical and physical properties of the butt-welded bars of AA6061-
T6 by a recently developed RAFSW technique using low-cost 3-axis machine tools.
• The joint efficiency of a naturally aged weld reaches 72% after 14 days.
• The optimized conditions for artificial aging are obtained as presented in Table 18. It is
found that the joint efficiency of RAFSW samples reaches 90% by artificial aging at 180 °C
for 18 h.
• Moreover, it is found that industrial users can take advantage of a fast artificial aging
process at 220 °C for 30 min to obtain a RAFSW joint with 85% of joint efficiency and 241
MPa of tensile strength.
• In industry, a delay is likely to happen between welding and PWHT. In this regard, the
repeatability of the optimized artificial aging processes is validated for the time gaps up to
21 days between the welding and PWHT process.
• Solubilizing heat treatment prior to artificial aging, called W+T6, at 200 °C for 2 h results
in improvement of the tensile strength of RAFSW samples up to 243 MPa. However, the
welds have low ductility compared to the samples just aged artificially.
• It is shown that the artificial aging process on RAFSW samples not only is more time and
cost-effective than the solubilizing followed by artificial aging process, but also yields
higher mechanical properties even when it is done at high paces.
The results provided in this research can be used as a guideline regarding the effect of various PWHT
processes on AA6061-T6 butt joints made by the recently developed RAFSW technique applicable
industrially on common low-cost 3-axis CNC machines.
Funding: This research has been supported by Team PI2 funds.
Acknowledgments: The authors would like to thanks the PI2/REGAL team members who made this
research possible.
Conflicts of Interest: The authors declare no conflict of interest.
85
Effect of Tool Design and Process
Parameters on Lap Joints Made by Right Angle
Friction Stir Welding (RAFSW)

Keywords: Friction stir welding; Lap joint; Tool design; Process parameters; Taguchi design of
experiments; Artificial neural network modeling
This Chapter has been published in: Journal of Manufacturing and Materials Processing, Vol 3 (3),
66 (2019)
https://doi.org/10.3390/jmmp3030066
Licensee MDPI, Basel, Switzerland. This is an open access article distributed under the Creative
Commons Attribution License which permits unrestricted use, distribution, and reproduction in any
medium, provided the original work is properly cited.
(https://creativecommons.org/licenses/by/4.0/)
86
5.1 Résumé
Au cours des dernières décennies, le soudage par friction-malaxage (FSW) a attiré l'attention des
secteurs universitaire et industriel en tant que développement le plus important dans les processus
d'assemblage des métaux. Le joint à recouvrement FSW est une alternative intéressante pour les
rivets, les soudures par fusion et le collage, en particulier dans l'industrie des transports. Dans cet
article, l'effet de la conception de l'outil et des paramètres de processus sur la force axiale descendante
générée et la résistance des joints à recouvrement AA6061-T6 est étudié. Les soudures sont réalisées
par une technique de soudage par friction-malaxage à angle droit à faible coût (RAFSW). Les
paramètres de conception d'outil étudiés sont le diamètre d'épaulement, la profondeur de rainure
d'épaulement, la longueur de broche, l'angle de broche, le diamètre de base de broche et le fil de
broche. De plus, l'effet de la vitesse de rotation de l'outil, de la vitesse de déplacement, de la
profondeur de plongée et de la configuration des joints de recouvrement est évalué. La méthode
Taguchi est utilisée pour concevoir les expériences et la modélisation du réseau neuronal artificiel
(ANN) est appliquée pour prédire la force de plongée et la résistance des articulations. Les résultats
indiquent qu'une soudure de qualité peut être obtenue à de faibles forces axiales vers le bas pendant
le soudage en sélectionnant correctement la conception de l'outil et les paramètres du processus. Il est
identifié que l'on peut obtenir un joint de recouvrement de qualité à des vitesses de déplacement aussi
élevées que 1400 mm / min et des forces axiales vers le bas aussi faibles que 3,2 kN par une technique
RAFSW à faible coût.
5.2 Abstract
In recent decades, friction stir welding (FSW) has attracted extensive attention of academic and
industrial sectors as the most considerable development in metal joining processes. FSW lap joint is
an interesting alternative for rivets, fusion welds and bonding particularly in transportation industry.
In this paper, the effect of tool design and process parameters on the generated downward axial force
and strength of AA6061-T6 lap joints is studied. The welds are made by a low-cost friction stir
welding technique at right angle (RAFSW). The studied tool design parameters are shoulder diameter,
shoulder groove depth, pin length, pin angle, pin base diameter, and pin lead. Moreover, the effect of
tool rotational speed, traverse speed, plunge depth, and lap joint configuration is evaluated. Taguchi
method is used to design the experiments and artificial neural network (ANN) modeling is applied
to predict the plunging force and the strength of the joints. The results indicate that a quality weld can
be obtained at low downward axial forces during welding by proper selection of tool design and
process parameters. It is identified that one can achieve a quality lap joint at traverse speeds as high
as 1400 mm/min and downward axial forces as low as 3.2 kN by a low-cost RAFSW technique.
87
5.3 Introduction
Friction stir welding (FSW) is a newly emerged solid-state joining process. In the past decades, it has
increasingly attracted the interest of researchers and industry owing to its prominent advantages over
fusion welding techniques. It is a green, versatile technology capable of making high quality welds.
However, FSW equipment costs and royalties limit the use of this technique in current industry [1, 2,
4, 19]. Some attempts are made to overcome this issue. In this regard, a low-cost friction stir welding
technique at right angle (RAFSW) has been developed by our research team, recently [29, 70]. It can
be employed on low-cost 3-axis CNC machining centers without any need to prior modification of
the machine. The RAFSW technique is capable of making sound, defect-free welds at zero tilt angle
with low axial forces during welding compare to common FSW techniques. Thus, this method not
only works with 3-axis machines instead of 4 or 5-axis machines, but also do not need sturdy, stiff,
and high capacity expensive equipment. Moreover, clamping and fixturing can easily use existing
vises and clamps of the CNC machining center since the required holding force is lower [29, 70].
Until now, FSW has primarily been employed to produce butt joints [71, 101] and sometimes for lap
joint. By developing FSW techniques for lap joints, the number of applications of this technique
would expand, drastically. Assembly of parts and components in the transportation industry is widely
done using lap joint configuration [60, 101, 102]. This configuration is extensively used in mechanical
structures in format of riveted joints and fusion welded joints. Taking advantage of FSW lap joints,
instead of the mentioned joints, can lead to reduction of weight, cost and production time of the joints.
Moreover, it boosts the mechanical properties of the joint since the joint has less defects and
imperfections than other types of joining processes [56, 58, 71].
Production of sound, quality lap joints is not with the same ease of making butt joints due to some
reasons [103, 104]. Firstly, in overlap joints, there are two crack-like unwelded zones that can act as
crack initiation sites when the joint is under load [58, 101, 103]. Moreover, there are two types of
defects in lap joints which are hooking and cold lap defects. Hook defect is not always the fracture
initiation site and sometimes is not really bad especially for dissimilar lap joints [105-107] but
sometimes, the fracture starts from theses defects [102, 103]. In general, they can have damaging
effect on the strength of the welds. Their adverse effect can be restricted and even avoided by proper
tool design and process parameters [58, 103, 104]. Furthermore, the disruption of the oxides at the
sheets interfaces is more difficult in the lap joint configuration [103]. In addition, the weld width
plays a considerable role in the joint performance [104]. The wider the width of the weld, the more
the downward axial forces generated during the welding process which is not desirable as it increases
the cost of equipment and fixturing. Moreover, plate thinning and entrapment of oxide particles
88
happen in the lap joints which must be minimized [58, 101, 103]. Additionally, in lap joints, there are
two overlapped sheets which tend to separate from each other when the FSW tool progresses in the
material. That can be avoided by more clamping and fixturing of the structure or by applying some
changes in the tool design compared to the tools for butt joints. The negative effects of the
aforementioned problems can be prevented by employing a proper tool design and adequate process
parameters [58, 60, 103]. Thus, the presence of these defects becomes negligible.
Although, there are numerous studies on the effect of tool design and process parameters [71, 102]
on the quality of the butt joints made by FSW technique, there is less research on the lap joints [101].
Some research have shown the effect of tool design including the shoulder shape (convex, flat, or
concave), shoulder dimension, pin shape (cylindrical or conical), and surface features on the shoulder
and pin (like flutes, grooves, or threads) on the quality of the lap joints [14, 58, 103]. For instance,
Yue et. al. have identified that a reverse-threaded pin works better than a threaded pin to make quality
AA2024 lap joints [108]. Buffa et. al. reported that the effective material flow greatly depends on the
tool design. Cylindrical-conical pins was the most effective design to make 2198-T4 lap joints in their
research [60]. Some other research have investigated the impact of the process parameters including
the tool plunge depth, rotational speed, traverse speed, and configuration of the lap joint on the quality
of the welds [56, 58, 109, 110]. For example, the effect of the process parameters on the mechanical
properties of AA5456 lap joints has been studied in a research. The results show that the optimal
mechanical properties are obtained when the rotational and traverse speed are 250 rpm and 75
mm/min, respectively [71]. In another study investigating the effect of process parameters on AA6060
lap joints, it was shown that the increase of the rotation speed causes the decrease of joint strength
[111]. In the majority of these studies, the traverse speed is too low for industrial purposes [71, 101,
110, 111]. Additionally, it was demonstrated that the lap joint configuration affects the strength of
the joints [58, 60] and that the best results were found when the advancing side of the weld is located
on the upper sheet of the lap joint [60].
Nowadays, there is an increasing interest to employ FSW techniques for lap joint assembly in the
transportation industry. AA6061-T6 like some other alloys has many examples of application such as
ship hulls, truck roof and side panels, wagon roofs. Unfortunately, there is a lack of information and
too few studies on the determination of the effective working window of the tool design parameters
and the process parameters to make quality AA6061-T6 lap joints by the FSW technique. To make
an extensive study on the effect of these parameters on the quality of joints, Taguchi method can be
employed to minimize the number of experiments [72]. Besides, artificial neural network modeling
can be used as a powerful tool to predict the behavior of the joints based on the experimental data
89
[73, 76]. These methods have successfully been applied in many researches regarding the FSW joints
mainly for butt joint configuration [70, 76, 101, 112].
In this paper, an extensive study is conducted on the effect of tool design parameters and process
parameters on the quality of AA6061-T6 lap joints made by RAFSW technique. To this end, Taguchi
method is used to design the experiments. Afterwards, ANN modeling is employed to predict the
effect of the mentioned parameters on the downward axial force during the welding process and the
strength of the joints represented in terms of fracture force. In this research, high traverse speeds are
applied to make the process promising for industrial use.
5.4 Materials and Methods

In this paper, RAFSW technique is employed to make the lap joints of AA6061-T6 extruded sheets
of 1.6 mm thickness. Table 20 presents the chemical composition and tensile strength of this sheet
metal according to the specification provided by the extruder. The sheets are cut to pieces with the
dimensions of 245 𝑚𝑚 × 88 𝑚𝑚 × 1.6 𝑚𝑚 using a shear press. A 3-axis CNC machining center,
Fryer MC-15, with 25 HP spindle using CAT40 tool holders is used in this research for friction stir
welding. The maximum rotational speed and the axis peak trust of this machine are 8000 rpm and 15
kN, respectively. To make the joints, two sheets are fixed on top of each other on a rigid back-plate
installed on top of a calibrated dynamometer. The dynamometer is a 3-axis Kistler 9265B. The
RAFSW tool is mounted into a long reach tool holder. In this paper, the specially designed tools to
make RAFSW lap joints are flat shoulder tools with some grooves on the shoulder. The pin is threaded
and have a conical shape. Single-pass welds are conducted along the extruded direction of the sheets.
Figure 67 depicts a general view of the tool to make lap joints by RAFSW, the RAFSW set-up,
clamping of the sheets on the Kistler dynamometer, the tool in the tool holder, and a welded sample.
To make the lap joints, there are two types of configurations, which are not of the same properties
due to the asymmetric nature of FSW joints. Thus, both types of configurations are studied in this
research as illustrated in Figure 68. The fracture force of the joints is investigated by the tensile shear
test. The weld coupons are machined with the dimensions specified in Figure 69. The top sheet was
loaded in the tensile shear test [56]. A hydraulic testing machine was used to conduct the uniaxial
tensile tests under a crosshead speed of 1 mm/min. The machine was equipped with a load cell of
44.5 kN calibrated to±0.08%. In this research, the single lap shear tests are done without spacers
because the goal of this paper is to establish the strength of the lap joints for the applications such as
assembly of truck panels, bus and wagon roofs. In such applications, the force applied to the joint is
not centered. Therefore, single lap shear test without spacers can replicate the applied forces closer
to reality than the test with spacers.
90
Table 20: Chemical composition and tensile strength of base metal.
Chemical Composition (wt%)
Material Al Mg Mn Cu Fe Si UTS (MPa)

AA-6061-T6 Bal. 0.83 0.07 0.19 0.19 0.55 285
Figure 67: (a) The tool shape to make lap joints by RAFSW technique, (b) the close view of the tool and tool
holder, (c) the RAFSW set-up including the tool, tool holder, clamped sheets on the back-plate, and the back-
plate installed on the dynamometer, (d) the welded sample, (e) the schematic of the tool and the design
parameters (the pin lead, not illustrated, is the distance between the threads on the pin).
91
Figure 68: The schematic of the welding process for two possible configurations of the lap joints.
Figure 69: the schematic and dimensions of the weld coupons to make the tensile test. (a) Configuration No.
1, when the advancing side is on the upper sheet (b) configuration No. 2, when the retreating side is on the
upper sheet.
5.5 Results and Discussion

5.5.1 Design of Experiments
The design of experiments (DOE) is performed based on Taguchi method. Accordingly, a L16
orthogonal array is used to explore the effect of the tool geometry and the process parameters on the
downward axial force and the fracture force of the lap joints made by RAFSW technique. The tool
geometry parameters are the shoulder diameter, the shoulder groove depth, the pin length, the pin
angle, the pin base diameter and the pin lead as shown in Figure 67. Like industrial tools, a 1 mm
radius is added to the edge of all tool shoulder to minimize burr formation and improve welding when
imperfect materials and assemblies are present. The process parameters are the tool traverse and
rotational speeds, the tool plunge depth and the lap joint configuration. Therefore, 16 experiments are
designed based on the L16 array. In addition, for each test of the array two passes named “a” and “b”
92
were done. In the first pass, the plunge depth is set exactly to the pin length. In the second pass, the
tool plunge depth is deeper by 0.05 mm and 0.08 mm for tool groove depths of 0.1 mm and 0.25 mm,
respectively. The designed L16 arrays and both sets experimental results are presented in Table 21.
In addition, some additional experiments are conducted to provide more data for the modeling
purposes, as shown in Table 22. The tools made for all experiments are illustrated in Figure 70.
Overall, 20 tools are made and 40 experiments are conducted in the experimental section of this
research.
Figure 70: (a) 16 tools made according to L16 DOE presented in Table 21, (b) 4 tools made regarding Table
22.
Table 21: The experiments conducted according to L16 orthogonal array. For each condition of the L16 array,
tow tests are done with different tool plunge depths.
Sampl Sampl PL SD SGD PBD P PLD V w PD C DA FF

e No. e Code (mm (mm (mm (mm A (mm (mm/min (rpm (mm F (N)
) ) ) ) (°) ) ) ) ) (N)
1 1-a 1.8 8.5 0.1 4 20 0.25 1400 3500 1.8 1 2600 299
4
2 1-b 1.8 8.5 0.1 4 20 0.25 1400 3500 1.85 1 3060 445
3
3 2-a 1.8 9.5 0.1 4 20 0.45 2000 5000 1.8 2 2500 339
0
4 2-b 1.8 9.5 0.1 4 20 0.45 2000 5000 1.85 2 3120 393
7
5 3-a 1.8 10.5 0.25 4.8 24 0.25 1400 3500 1.8 2 2900 537
3
6 3-b 1.8 10.5 0.25 4.8 24 0.25 1400 3500 1.88 2 4130 542
2
7 4-a 1.8 11.5 0.25 4.8 24 0.45 2000 5000 1.8 1 2810 284
7
8 4-b 1.8 11.5 0.25 4.8 24 0.45 2000 5000 1.88 1 4320 521
8
9 5-a 2.2 8.5 0.1 4.8 24 0.25 2000 5000 2.2 2 3210 351
0
10 5-b 2.2 8.5 0.1 4.8 24 0.25 2000 5000 2.25 2 3590 339
0
93
11 6-a 2.2 9.5 0.1 4.8 24 0.45 1400 3500 2.2 1 3420 521
8
12 6-b 2.2 9.5 0.1 4.8 24 0.45 1400 3500 2.25 1 4000 560
9
13 7-a 2.2 10.5 0.25 4 20 0.25 2000 5000 2.2 1 3200 450
2
14 7-b 2.2 10.5 0.25 4 20 0.25 2000 5000 2.28 1 3840 575
6
15 8-a 2.2 11.5 0.25 4 20 0.45 1400 3500 2.2 2 3310 270
9
16 8-b 2.2 11.5 0.25 4 20 0.45 1400 3500 2.28 2 4530 329
2
17 9-a 2.6 8.5 0.25 4 24 0.45 1400 5000 2.6 2 2970 408
3
18 9-b 2.6 8.5 0.25 4 24 0.45 1400 5000 2.68 2 3240 331
4
19 10-a 2.6 9.5 0.25 4 24 0.25 2000 3500 2.6 1 4100 335
8
20 10-b 2.6 9.5 0.25 4 24 0.25 2000 3500 2.68 1 4720 474
6
21 11-a 2.6 10.5 0.1 4.8 20 0.45 1400 5000 2.6 1 3170 610
7
22 11-b 2.6 10.5 0.1 4.8 20 0.45 1400 5000 2.65 1 3780 564
0
23 12-a 2.6 11.5 0.1 4.8 20 0.25 2000 3500 2.6 2 4480 217
1
24 12-b 2.6 11.5 0.1 4.8 20 0.25 2000 3500 2.65 2 4700 230
0
25 13-a 3 8.5 0.25 4.8 20 0.45 2000 3500 3 1 4320 498
2
26 13-b 3 8.5 0.25 4.8 20 0.45 2000 3500 3.08 1 4680 655
7
27 14-a 3 9.5 0.25 4.8 20 0.25 1400 5000 3 2 2930 212
3
28 14-b 3 9.5 0.25 4.8 20 0.25 1400 5000 3.08 2 3360 204
2
29 15-a 3 10.5 0.1 4 24 0.45 2000 3500 3 2 3934 213
6
30 15-b 3 10.5 0.1 4 24 0.45 2000 3500 3.05 2 4390 205
6
31 16-a 3 11.5 0.1 4 24 0.25 1400 5000 3 1 3040 358
1
32 16-b 3 11.5 0.1 4 24 0.25 1400 5000 3.05 1 3750 414
1
(Abbreviation description: Pin length (PL), Shoulder Diameter (SD), Shoulder Groove Depth (SGD), Pin base diameter
(PBD), Pin angle (PA), Pin lead (PLD), Traverse speed (V), Rotation speed (w), Plunge Depth (PD), Configuration (C),
Downward axial force (DAF), Failure force (FF))
94
Table 22: Some more experiments conducted to explore more regarding the effect of tool and process
parameters on downward axial force and fracture force.
Sampl Sampl PL SD SGD PBD P PLD V w PD C DA FF

e No. e Code (mm (mm (mm (mm A (mm (mm/min (rpm (mm F (kN)
) ) ) ) (°) ) ) ) ) (kN)
33 17-a 1.8 8.5 0.1 3.5 20 0.25 1700 4250 1.8 1 2970 311
8
34 17-b 1.8 8.5 0.1 3.5 20 0.25 1700 4250 1.835 1 3320 398
2
35 18-a 1.75 12 0.3 5 25 0.5 1800 4500 1.75 2 2760 545
4
36 18-b 1.75 12 0.3 5 25 0.5 1800 4500 1.83 2 4320 596
1
37 19-a 1.75 12 0.3 4.5 25 0.35 2200 5000 1.75 1 3340 222
4
38 19-b 1.75 12 0.3 4.5 25 0.35 2200 5000 1.83 1 4600 404
3
39 20-a 1.75 11.5 0.3 4.5 25 0.45 2400 5500 1.75 2 3020 432
8
40 20-b 1.75 11.5 0.3 4.5 25 0.45 2400 5500 1.83 2 4000 458
2
5.5.2 Artificial Neural Network Modeling

The experimental data presented in the previous section were utilized to train artificial neural
networks in this section. Feed-forward neural networks with backpropagation algorithm were used to
model the relationship between the tool geometry and the process parameters on the downward axial
force and the fracture force of the joints. Indeed, several factors affect the prediction accuracy of the
backpropagation neural network models such as architecture of the network, momentum coefficient,
and learning rate of the model [73, 74]. Accordingly, in this paper, the effect of these parameters on
the accuracy of the neural network models is studied to find the best ANN modeling factors. An ANN
model with too small architecture can results in insufficient degree of freedom; and too large network
causes to over fit the data. Thus, there is an optimized number of neurons and hidden layers to have
a reliable ANN model [75]. In this research, several number of neurons in one hidden layer were
evaluated to find the optimal architecture. The studied factors to compare the different architectures
were root-mean squared error (RMSE), maximum error, mean relative error (MRE), and mean
absolute error (MAE) [76]. Moreover, the effect of the learning rate and the momentum coefficient
in the accuracy of the models is studied. The correlation of the learning rate, the momentum
coefficient with RMSE, and maximum error are illustrated in Figure 71. To keep the RMSE and
maximum error minimized, the optimal learning rate and momentum coefficient is 0.5 for both of
95
them. Additionally, it was found that the optimal architecture is 10-8-1 for both models of downward
axial forces and fracture forces shown in Figure 72. Table 23 presents the employed condition to
make the final ANN models in this paper.
Figure 71: The effect of learning rate and momentum coefficient on (a) root-mean squared error (RMSE) and
(b) maximum error when the architecture of the model is 10-8-1.
Figure 72: The architecture of ANN models in this paper. The architecture of the model for downward axial
force and fracture force is 10-8-1 for both of them.
96
Table 23: The details regarding the developed ANN models in this paper to study the effect of tool geometry
and process parameters on downward axial force and failure force.
Welding Force model Failure Force model

Network configuration 10-8-1 10-8-1
Number of inputs 10 10
Number of hidden layers 1 1
Number of neurons 8 8
Number of outputs 1 1
Total No. of experimental data 40 40
Learning rate 0.5
Momentum 0.5
The experimental data of the 40 tests obtained in the previous section were used to develop the ANN
models, in this section. 36 experiments are separated from 40 experiments as training data for the
ANN models; and the 4 remaining tests were utilized as confirmation tests to evaluate the
predictability and the accuracy of the developed ANN models. Table 24 shows the RMSE and
maximum error of the developed 10-8-1 ANN models for training data, confirmation data, and the
entire set of the data. It is important to keep the error of the both trained data and confirmation data
minimized. The RMSE, MRE, MAE, and maximum error of the entire set of experiments for the
developed ANN models are presented in Table 25. From Table 24 and Table 25, it is concluded that
the amount of different kinds of errors for the developed models are low enough to make them capable
of accurate predictability. Furthermore, the relationship between the experimental data and predicted
values by the developed ANN models is studied by calculating the amount of correlation coefficient
(R2) using linear regression analysis. When the correlation coefficient is close to one, it indicates a
close relationship between the experimental data and predicted data by the developed ANN models
[73, 76, 77]. According to Table 26, the correlation coefficient for training data, confirmation data,
and the overall data are respectively 0.999981092, 0.998622385, and 0.999846253 for the ANN
model of downward axial force. For the failure force model, they are 0.999722461, 0.998734858, and
0.999634318, respectively. These coefficient values confirm the accuracy of both downward axial
force and failure force models. The comparison between experimental and predicted values, shown
in Table 27, indicate errors of less than 5% in all cases. In summary, Table 24 to Table 27 validate
the reliability of the developed ANN models to show the effect of the tool geometry and the process
parameters of the downward axial force and the fracture force of the lap joints made by RAFSW
technique.
97
Table 24: The RMSE and maximum error for the training data, the confirmation data, and the overall data
when the architecture of the neural networks is 10-8-1.
Training Data Confirmation Data All Data

Downward axial Failure Force Downward axial Failure Force Downward axial Failure Force
Force Force Force
RMSE Max RMSE Max RMSE Max RMSE Max RMSE Max RMSE Max
Error Error Error Error Error Error
15.97 52 71.49 211 135.76 169 129.53 174 45.53 169 80.96 211
Table 25: The amount of different kind of errors for the developed ANN models.
RSME MAE MRE Maximum error

1⁄ 𝑁 𝑁
𝑁 2 1 1 |𝐴𝑖 − 𝑌𝑖 |
formula 1 |𝐴𝑖 − 𝑌𝑖 |
( ∑(𝐴𝑖 − 𝑌𝑖 )2 ) ∑|𝐴𝑖 − 𝑌𝑖 | ∑( ) × 100
𝑁 𝑁 𝑁 𝐴𝑖
𝑖 𝑖
𝑖
Downward axial Force model 45.53 21.61 0.64% 169
Failure Force model 80.96 61.85 1.73% 211
The formulas are from Ref.[76])
Table 26: The amount of correlation coefficient (R2) for the training data, confirmation data, and overall data
for the developed ANN models.
∑𝑵
𝟐
R2 R2 R2
𝑹𝟐 = 𝟏 − ( 𝑵 𝟐
)
∑𝒊 (𝒀𝒊 )
(For training data) (For confirmation tests) (For all data)
Downward axial Force model 0.999981092 0.998622385 0.999846253
Failure Force model 0.999722461 0.998734858 0.999634318
The formula is from Ref.[76])
Table 27: The amount of experimental data, predicted data, and its error for the confirmation experiments.
Sample Measured Predicted Error of model Measured Predicted Error of

No. downward downward for downward failure force failure force model for
axial Force (N) axial Force (N) axial Force (N) (N) (N) failure
force (N)
15 3310 3479 5% 2709 2810 3.7%
20 4720 4553 3.5% 4746 4876 2.7%
27 2930 3050 4.1% 2123 2025 4.6%
34 3320 3373 1.6% 3982 4156 4.4%
|Ai −Yi |
(Error of model = Ai
× 100)
5.5.3 Effect of Tool Geometry Parameters on Welding Force and Tensile Shear Force
In this section, the effect of the tool geometry parameters on the downward axial force during RAFSW
process and the fracture force at tensile shear test is studied, according to the developed ANN models
in the previous section. Generally, a compromise between the fracture force and the downward axial
force yields the best condition to make the welds by RAFSW technique. The reason is that the
mentioned condition makes that possible to use low capacity, cost-effective CNC machines with
98
minimized clamping and fixturing that causes to have a low-cost RAFSW process [70]. Therefore,
having a joint with high fracture force while keeping the downward axial force minimized would be
desirable. Among the designed and tested tools, tool No. 21 meets this condition. Therefore, the
following studies are done around this condition.
Figure 73 depicts the correlation of the shoulder diameter and the shoulder groove depth with the
downward axial force and the fracture force. The downward axial force significantly increases by the
increase of the shoulder diameter as shown in Figure 73.a. It is due to the increase of shear,
deformation, and material friction during the welding process [103]. Therefore, higher axial forces
are generated during the welding process. Moreover, the increase of the shoulder groove depth causes
to the slight increase of the downward axial force. This could be related to the increase of the amount
of the material involved in the mixing process between the grooves of the tool and the workpiece
material. This leads to increase of the friction between the tool and material, slightly [103]. According
to Figure 73.b, the impact of the shoulder groove depth on the fracture force is negligible, while the
increase of the tool shoulder diameter causes the decrease of the fracture force. It could be associated
to the fact that at higher shoulder diameters, the downward axial force is higher which causes a higher
heat input. Thus, the heat-affected zone, HAZ, around the nugget zone will be larger [14]. As the
HAZ is the weakest place in the weld area in terms of strength, the larger HAZ causes the lower
strength of the joint. Therefore, the fracture force decreases.
Figure 73: Contour plots of: (a) downward axial force during RAFSW process, and (b) failure force at tensile
shear test versus shoulder diameter, SD, and shoulder groove depth, SGD, while other geometry and process
parameters are the same as parameters of sample No. 21. The contour plots are extracted from developed
ANN models in this paper.
As shown in Figure 74.a, the increase of the pin length leads to the increase of the downward axial
force. It is attributed to the higher friction between the tool pin and the workpiece material during the
mixing and stirring process due to larger interfacial area [14, 109]. When the pin angle increase from
99
19° to 21°, the downward axial force increases at lower pin lengths. However, the axial force
decreases by increase of pin angle at higher pin lengths. This behavior could be assigned to the
interwoven relationship between the effect of pin angle and pin length on the material flow during
the welding. Thus, it has a complex effect on the stirring and mixing mechanisms during the welding
process. Figure 74.b illustrates the correlation of the pin length, the pin angle, and the fracture force.
It indicates that the fracture force gradually increases when the pin angle increases from 19° to 21°.
It could be due to the enhancement of the efficiency of stirring mechanism [14]. Moreover, the
increase of the pin length leads to the decrease of the fracture force. It can be due to the formation of
more deteriorating hooking and cold lap defects. In fact, hooking and cold lap defects are the main
cause of weakness of the FSW lap joints. The geometry and severity of these defects are affected by
tool geometry and process parameters, considerably [14, 109].
Figure 75 illustrates the impact of the pin base diameter and the pin lead on the the downward axial
force and the fracture force. It can be seen that the larger the pin base and the pin lead, the higher the
downward axial force, as shown in Figure 75.a. The reason is that the friction between the tool and
workpiece increases by increase of these parameters that eventually causes to increase of the axial
force [14]. According to Figure 75.b, increase of the pin base diameter and the pin lead result in
increase of the fracture force. It can be attributed to the fact that at constant pin length and plunge
depth, the increase of the pin base diameter and the pin lead can causes to formation of less defects
in the joint. As a result, the fracture force boosts by these changes [14].
shear test versus pin length, PL, and pin angle, PA, while other geometry and process parameters are the same
as parameters of sample No. 21 except for the plunge depth wich is the same as pin length. The contour plots
are extracted from developed ANN models in this paper.
100
shear test versus pin base diameter, PBD, and pin lead, PLD, while other geometry and process parameters are
the same as parameters of sample No. 21. The contour plots are extracted from developed ANN models in this
paper.
5.5.4 Effect of Process Parameters on Welding Force and Tensile Shear Force
The correlation of the welding tool traverse speed and the tool rotational speed with the downward
axial force is depicted in Figure 76.a. It can be observed that the increase of the traverse speed leads
to an increase of the downward axial force. It is due to the fact that the more the traverse speed, the
material in front of the tool would be colder [1]. Indeed, the higher the tool traverse speed, the less
heat input would be generated in the weld area [1]. Thus, the needed axial force to pass the tool
through the cold material of the workpiece would be high. In addition, Figure 76.a indicates that the
higher the rotational speed, the lower the downward axial force. This is due to the increase of the heat
input at higher rotational speed that causes softer and warmer material ahead of the tool [1]. Figure
76.b illustrates that the increase of the tool rotational speed causes lower failure force. It could be due
to the excessive heat input at high rotational speeds. The excessive heat input makes the HAZ area
wider which has an adverse effect on the strength of the joint [1]. Moreover, this figure shows that
the increase of the tool traverse speed reduces the failure force by affecting the mixing and stirring
processes, adversely. This results in the formation of defects and weak joints [56, 110]. Compare to
butt joints, lap joints strength highly depends on hook defects which does not exist in butt joints. In
butt joints, the fracture happens in low hardness region which is usually located in the HAZ area
[113], while besides the weakness in the hardness of HAZ area in lap joints, other factors such as
hook defect, cold lap defect, and tow crack-like sites play important role in fracture of the lap joints.
As a result, differences in the strength of the butt joint and lap joints can be observed in response to
the variation of tool design and process parameters such as traverse speed [113].
101
shear test versus welding traverse speed, V, and welding rotational speed, w, while other geometry and
process parameters are the same as parameters of sample No. 21. The contour plots are extracted from
developed ANN models in this paper.
According to Figure 77.a, the downward axial force increases with the plunge depth at a given tool
traverse speed (when the pin length is the same as the plunge depth). Additionally, as shown in Figure
78.a, a higher plunge depth increases the downward axial force at a given rotational speed. These are
due to the higher frictional contact between the tool and the material at higher plunge depths [14,
109]. Figure 77.b and Figure 78.b also show that higher plunge depths lower the fracture force which
can be attributed to the formation of more damaging hooking defect and a large extent of upper sheet
thinning in the weld area when the tool plunge depth is too high [56, 102, 111].
Generally, in both experimental results and ANN models, configuration 1 (when the advancing side
is on the upper sheet) is preferable in terms of the failure force. This is in conjunction with reported
research [58, 60]. This is mainly due to the fact that the cold lap defect is not as deteriorating as
hooking defect [102]. When the upper sheet is on advancing side, the cold lap defect is present in this
side. Additionally, in all experiments, the fracture occurred in the welded region of the joint. This
means that the reason of variation of joint strength is really related to the tool design and process
parameters used for each weld.
As mentioned, a compromise between the fracture force and the downward axial force yields the best
condition to make the welds by RAFSW technique. Therefore, having high fracture force while
keeping the downward axial force minimized would be desirable. Based on Figure 73 to Figure 78,
an efficient working window of the tool geometry and process parameters to make lap joints by
RAFSW technique is presented in Table 28.
102
shear test versus tool plunge depth, PD, and welding traverse speed, V, while other geometry and process
parameters are the same as parameters of sample No. 21 except for the plunge depth which is the same as pin
length. The contour plots are extracted from developed ANN models in this paper.
shear test versus tool plunge depth, PD, and welding rotational speed, w, while other geometry and process
parameters are the same as parameters of sample No. 21 except for the plunge depth wich is the same as pin
length. The contour plots are extracted from developed ANN models in this paper.
Table 28: An efficient working window of the geometry and process parameters to make lap joints by
RAFSW.
PL SD SGD PBD PA PLD V W PD C

(mm) (mm) (mm) (mm) (°) (mm) (mm/min) (rpm) (mm)
2.5- 9.5- 0.09-0.12 4.7-4.9 19- 0.44- 1200- 4500- 2.5- 1
2.65 10.5 20 0.46 1450 5200 2.65
103
5.6 Conclusions
In the present study, RAFSW technique was applied to make AA6061-T6 lap joints. The main goal
was to evaluate the effect of the tool design and the process parameters on the downward axial force
generated during the welding process and the fracture force of the joints. Taguchi method is used to
minimize the number of experiments; and ANN modeling is implemented to anticipate the behavior
of the joints. The main results of this study are presented as follows (all the statements are related to
the studied range of the parameters in this paper):
• The effect of tool design and process parameters on the quality of the lap joints and the
generated forces during the RAFSW process have been predicted with high accuracy using
ANN modeling.
• Larger shoulder diameters and pin lengths cause an increase of the downward axial force and
a decrease of the fracture force of the joints.
• The downward axial force and the fracture force of the joints increase with an increase of pin
base diameter and pin lead. However, the variation of shoulder groove depth and pin angle
has minor effects on the axial force and the failure force, in the studied range.
• The increase of the tool rotational speed causes to reduce of the downward axial force and
the fracture force. The downward axial force increases and the fracture force decreases with
an elevation of tool traverse speed and rotational speed.
• The efficient range for tool design and process parameters to make quality lap joints at good
traverse speeds is provided in Table 28.
• Making quality lap joints at high traverse speed while keeping the downward axial force as
low as possible is the most promising condition for industrial users. According to this study,
one can accomplish a sound, quality AA6061-T6 lap joint made by RAFSW technique at
traverse speeds as high as 1400 mm/min and downward axial forces as low as 3.2 kN.
The results of this study can be used as a roadmap to make quality AA6061-T6 lap joints by RAFSW
technique for industrial applications.
Funding: This research has been supported by funds of PI2 Team.
Acknowledgments: The authors would like to thanks the PI2/REGAL team members who made this
research possible.
Conflicts of Interest: The authors declare no conflict of interest.
104
Implementation of Right Angle Friction Stir
Welding (RAFSW) to Assemble the Side Panels of
Truck Box

Keywords: Friction stir welding; Force control; Truck side panel; Taguchi method; Artificial neural
network modeling; Fatigue test
This Chapter has been accepted to publish in: The International Journal of Advanced Manufacturing
Technology, July 2020.
105
6.1 Résumé
Le soudage par friction-malaxage (FSW) a attiré de nombreuses attentions dans les secteurs
académiques et industriels en raison de ses caractéristiques prometteuses par rapport aux autres
techniques d'assemblage conventionnelles. Cependant, il a encore besoin de quelques améliorations
pour devenir fiable et rentable pour les applications industrielles. Dans cette étude, l'objectif est de
développer une technique de soudage par friction-malaxage à angle droit (RAFSW) pour assembler
de grands panneaux latéraux de caisses de camion afin que la technique soit fiable et rentable sur le
plan industriel. À cette fin, la méthode Taguchi et la modélisation du réseau neuronal artificiel (ANN)
ont été utilisées pour concevoir les expériences et optimiser les paramètres du processus de soudage.
La fenêtre de travail efficace des paramètres du procédé de soudage a été obtenue à des vitesses de
déplacement pouvant atteindre 1700 mm/min. Ensuite, la technique a été développée pour fonctionner
en mode de contrôle de force en utilisant un porte-outil à ressort et un outil nouvellement conçu avec
une entrée chanfreinée. Il est démontré que la technique développée pourrait tolérer des niveaux
élevés de perturbations de l'assemblage des joints comme de grands écarts entre les tôles avant le
soudage. De plus, il a été démontré que la technique développée pouvait être utilisée à l'aide de grands
routeurs d'usinage CNC à faible coût et avec un serrage uniquement aux coins des panneaux.
L'utilisation d'étiquettes de soudure avant le soudage a été bénéfique afin de réduire le serrage et la
distorsion tout en maintenant un niveau de résistance élevé le long des joints. Le logiciel par éléments
finis NX-NASTRAN et diverses expériences, y compris des tests statiques et de fatigue des panneaux,
ont été utilisés pour concevoir et évaluer l'assemblage soudé. Il a été constaté que la technique
RAFSW développée peut être utilisée pour l'assemblage de panneaux de camion.
6.2 Abstract
Friction stir welding (FSW) has attracted many attentions in both academic and industrial sectors due
to its promising characteristics compared to other conventional joining techniques. However, it still
needs some improvements to become reliable and cost-effective for industrial applications. In this
study, the aim is to develop a right angle friction stir welding (RAFSW) technique to assemble large
side panels of truck boxes so that the technique be industrially reliable and cost-effective. For this
purpose, Taguchi method and artificial neural network (ANN) modeling were used to design the
experiments and optimize the welding process parameters. The efficient working window of welding
process parameters were obtained at traverse speeds as high as 1700 mm/min. Then the technique
was developed to operate in force control mode using a spring load tool holder and a newly designed
tool with a tapered entry. It is shown that the developed technique could tolerate high levels of joint
fit-up disturbances like big gaps between the sheets before welding. Moreover, it was demonstrated
106
that the developed technique could be operated using large, low cost, CNC machining routers and
with clamping only at the corners of panels. The use of weld tags prior to welding was beneficial in
order to reduce clamping and distortion while maintaining a high strength level along the joints. NX-
NASTRAN Finite Element software and various experiments including static and fatigue tests of the
panels were used to design and evaluate the welded assembly. It was found that the developed
RAFSW technique has the potential to be used for truck panel assembly.
6.3 Introduction
6.3.1 Friction stir welding
Friction stir welding (FSW) is a solid-state joining process that has many advantages compared to
other metal joining techniques such as fusion welding and riveting [1, 57]. The FSW has great
potential to replace other assembly techniques in the industry [19]. In the last decade, an increasing
number of researches have been devoted to developing the FSW technique, notably to make lap joints,
which are widely used to assemble components and structures, especially in the transportation
industry. For instance, truck box, ship decks, railway tankers, and wagon roofs are typical examples
containing panels straightened with stringers and profiles and using lap or T joints [57, 60].
There are various methods to assemble side panels such as riveting, fusion welding, bonding, and
FSW. Riveting is time-consuming and unaesthetic; and the strength of the joint is low. However, it
does not need high labor skills, and is a good choice for unweldable aluminum alloys. Bonding is
less time-consuming, aesthetic, and affordable but it needs many jigs and clamps to fix the parts
during production. In addition, industries hesitate to use bonding due to its unknown long-term
durability. Fusion welding techniques such as MIG or TIG welding are mostly applicable to thicker
metal sheets and extrusions. In addition, they are slow, expensive and somewhat unaesthetic.
Furthermore, they may cause part distortion. Resistance spot welding (RSW) is limited in strength
but it is inexpensive. Finally, the FSW technique is not a cost-effective method to assemble parts in
most applications, yet. Whereas the joint strength is high, the distortion is rather low. The time spent
to assemble the panels is lower than riveting; and the outside appearance is nice-looking [56-58]. It
is also reported that FSW lap joints demonstrate a higher shear failure strength than conventional
riveted joints [114]. Recently, the automotive industry has been interested to replace riveted joints
by FSW lap joints.
To make the FSW process cost-effective for industrial users, PI2/REGAL team at Laval University
has recently developed FSW techniques operated only at right angle (RAFSW) [29, 70]. In these
techniques, there is no need to modify CNC machining centers or adapt them for the welding process.
107
One can make sound, quality welds at rather low forces by these techniques. In addition, this
technique can be implemented on low-cost 3-axis machines that are not sturdy, stiff, or of high
capacity. So far, the RAFSW technique has been developed to make butt-joints [70] and lap joints
[115] in position control mode.
6.3.2 Truck box side panels

Typically, the truck box sidewalls include thin metal sheets and stiffeners, as illustrated in Figure 79.
The panel must be rigid and dimensionally stable under the working loads. In the panel, the stiffeners
may have various designs like Z or delta shapes. The design selected for this paper is the delta, as
shown in Figure 79.b and c. The side walls of the truck panels also include holes in the stiffeners to
anchor the merchandise. The loads applied to the side panels come from the tension in the fastening
belt that holds the merchandise in place and from the truck box deformations during the truck motion.
Generally, the joints are the weakest regions of a panel, so, it is critical to assure their strength under
the applied load cases.
(b)
(a)
(c)
Figure 79: Illustration of the application: (a) truck box, (b) side panels, (c) dimension of the delta-shaped
stiffener.
6.3.3 Objectives of this research

Based on the previous RAFSW experiments of the lap joints of 1.6-mm-thick AA6061-T6 plates, it
has been possible to optimize the tool design and some operating conditions [115]. In this paper, it
is intended to develop and adapt the technique for the industrial production of side panels. To
108
implement the technique in production, the welding process must be re-optimized to improve
productivity and to be more tolerant to geometric errors in the assembly. In addition, the strength of
the welded parts must be maximized to sustain the load under different cases. Furthermore, the panel
distortions induced by the heat and the forces generated during the welding process must be
minimized.
Accordingly, this paper presents the main steps of a progressive approach that includes: (1) a new
experimental optimization of lap joints using the RAFSW technique in position control by a 3 axis
CNC machining center. (2) The adaptation of the process to become more tolerant to the geometric
errors including the RAFSW of small delta-shaped assemblies using the CNC machining center and
new tool designs and a new spring loaded tool holder. (3) Application and tests of larger side panel
assemblies.
6.4 Material
The truck box is built mainly from AA6061-T6 sheets and extrusions. Consequently, for all
simulations and tests, AA6061-T6 1.6 mm thick sheets are used to make both the skin and the stiffener
of the truck panels. Table 29 indicates the chemical composition and tensile strength of the sheet
metals used in this paper. The ultimate tensile strength (UTS) as received was found to be 285 MPa.
Table 29: Chemical composition and tensile strength of the sheet metal.
Chemical Composition (wt%)

Al Mg Mn Cu Fe Si UTS (MPa)
AA-6061-T6 Bal. 0.83 0.07 0.19 0.19 0.55 285
6.5 Optimization of lap joint welding

In this section, the RAFSW experiments are realized and modeled using neural networks to find the
range of parameters that can industrially be applied to maximize the joint strength and minimize the
downward axial force. The process parameters involved are the tool plunge depth, rotational and
traverse speeds, and the overlap length of the two sheets.
6.5.1 Apparatus
The RAFSW experiments are done using a 3-axis CNC machining center, Fryer MC-15, with 25 HP
spindle using a standard CAT40 tool holders. Position control is applied by default on CNC
machining centers. A shear press has been used to cut the 1.6 mm thick sheets to the dimensions of
245 mm × 88 mm. For each welding test, a pair of two sheets in lap joint configuration is fixed on
a rigid back-plate installed on the top of a 3-axis Kistler 9265B dynamometer. The tool design
109
parameters, presented in Table 30, have previously been optimized for RAFSW of lap joints [115].
The tool is mounted into a long reach tool holder. Single-pass welds are conducted along the rolled
direction of the sheets. Figure 80.a and b illustrate the used tool and tool holder, respectively. The
entire welding set-up and a sample welded by this technique are shown in Figure 80.c and d,
respectively. From the welded assemblies, 20 mm wide samples are cut, machined on their sides,
and tested in traction to find the failure force (FF). Figure 81 demonstrates the dimensions of the
samples.
Table 30: The characteristics of the optimized tool used in this research.
Shoulder Shoulder groove Shoulder edge Pin length Pin angle Pin lead Pin base diameter
diameter (mm) depth (mm) radius (mm) (mm) (°) (mm) (mm)
11.5 0.1 1 2.6 20 0.25 4.8
Figure 80: (a) The tool to make lap joints by RAFSW, (b) The close view of the tool and tool holder, (c) The
RAFSW set-up including the tool, tool holder, clamped sheets on the back-plate and the back-plate installed
on the dynamometer, (d) The welded sample [1].
Figure 81: The schematic and dimensions of the weld coupons to make the tensile test.
110
6.5.2 Design of experiments
To make the experimentation more efficient, a L9 orthogonal array based on the Taguchi Method is
used to explore the effect of process parameters on the downward axial force, DAF, and the failure
force, FF, of the lap joints. The process parameters are the tool traverse speed, V, the tool rotational
speed, w, the tool plunge depth, PD, and the overlap length of the two sheets, OL. The L9 array and
the experimental results are presented in Table 31. The study was completed with some additional
experiments to provide more data for interaction modeling purposes, as shown in Table 32.
Confirming tests are done to verify the predictability of the developed ANN models, as presented in
Table 33. Finally, a final test has been carried out after optimization based on the ANN model, as
presented in Table 34. Altogether, 18 experiments are conducted in this section. In all experiments,
the failure occurred in the weld area during the tensile test.
Table 31: The experiments conducted according to the L9 orthogonal array.
Sample's Code V (mm/min) w (rpm) PD (mm) OL (mm) DAF (N) FF (N)

L1 1300 3500 2.6 11.8 2930 7661
L2 1300 4250 2.65 15.6 3630 7261
L3 1300 5000 2.7 18.8 4040 6063
L4 1600 3500 2.65 18.9 3950 7373
L5 1600 4250 2.7 11.6 3860 6374
L6 1600 5000 2.6 15.6 2930 5848
L7 2000 3500 2.7 15.6 4700 7346
L8 2000 4250 2.6 18.7 3650 6954
L9 2000 5000 2.65 11.1 3590 5297
Table 32: The additional tests.

AT1 1300 5000 2.64 18.7 3630 6612
AT2 2000 3500 2.677 15.6 4440 6954
AT3 2000 4250 2.56 18.4 3150 4864
AT4 2100 4800 2.615 17.2 3490 5811
Table 33: The confirming tests.

CT1 1850 4850 2.67 17.7 3670 6812
CT2 2050 5000 2.68 17.2 3920 7066
CT3 1300 3500 2.6 14.3 3280 7660
CT4 1600 4250 2.7 15.5 4220 7653
Table 34: The optimized experiment.

Opt1 1700 4400 2.65 16.8 3650 7780
111
6.5.3 Artificial Neural Network Modeling
Two multi-layer perceptron models are used to predict the effect of process parameters on the
downward axial force and failure force of the lap joints. These ANN models are trained using the data
of Table 31 and Table 32. The data of Table 33 and Table 34 are utilized as confirmation tests to
evaluate the accuracy and predictability of the developed models. Network training applies the
backpropagation algorithm using momentum and learning rate of 0.5. From the modeling errors, it
was found that the optimal architecture is 4-9-1, i.e. 4 inputs, 9 neurons in the hidden layer and 1
output neuron. Table 35 and Figure 82 illustrates the characteristics of the optimal ANN models. The
optimal architecture has been found based on keeping the different types of modeling errors
minimized. Table 36 indicates the root-mean squared error (RMSE) and maximum error for the
training data, the confirming tests and the overall data. The RMSE, mean relative error (MRE), mean
absolute error (MAE), and maximum error of the entire set of experiments for the developed ANN
models are shown in Table 37. According to Table 36 and Table 37, the errors of the developed ANN
models are low enough to have good predictability. In addition, the amount of correlation coefficient
(R2) is calculated to evaluate the relationship between the experimental and predicted data. As shown
in Table 38, the R2 values for training, confirming, and overall data are 0.999938295, 0.998425175,
and 0.999508986, respectively, for the DAF model and 0.999658564, 0.998359221, and
0.999253618, for the FF model. A value of 1 indicates a perfect match of the experimental and
predicted data. Thus, the amount of the mentioned R2 verifies the accuracy of the developed models
for downward axial and failure force. The experimental and predicted values of confirming tests are
shown in Table 39. It can be observed that the predicting error is less than 5%. In brief, the data
presented in Table 36 to Table 39 confirm the accuracy and predictability of the developed ANN
models.
Table 35: The details of the developed ANN models for both the downward axial force and failure force.
Characteristics of the models DAF and FF Models

Network configuration 4-9-1
Number of inputs 4
Number of hidden layers 1
Number of neurons 9
Number of outputs 1
Total no. of experiments 18
Learning rate 0.5
Momentum 0.5
112
Figure 82: The architecture of developed ANN models for the downward axial force and fracture force
(4-9-1).
Table 36: The RMSE and maximum error for the training, confirmation, and entire set of data
Training Data Confirmation Data All Data
DAF FF DAF FF DAF FF
RMSE Max E. RMSE Max E. RMSE Max E. RMSE Max E. RMSE Max E. RMSE Max E.
29 69 123 270 150 192 292 388 83 192 186 388
Table 37: The amount of different kinds of errors for the final developed ANN models.
RSME MAE MRE Maximum Error

1⁄ 𝑁 𝑁
𝑁 2
1 1 1 |𝐴𝑖 − 𝑌𝑖 |
Formula ( ∑(𝐴𝑖 − 𝑌𝑖 )2 ) ∑|𝐴𝑖 − 𝑌𝑖 | ∑( ) × 100 |𝐴𝑖 − 𝑌𝑖 |
𝑁 𝑁 𝑁 𝐴𝑖
𝑖 𝑖 𝑖
DAF model 83 55 1.4% 192

FF model 186 139 2% 388
(Ai, Yi, N and i are the experimental value, predicted value, total number of experimental data and trial number, respectively.
The formulas are from [76])
Table 38: The amount of correlation coefficient (R2) for the training, confirmation and entire set of data.
∑𝑵
𝟐
R2 R2 R2
𝑹𝟐 = 𝟏 − ( )
∑𝑵𝒊 (𝒀𝒊 )
𝟐 (For Training Data) (For Confirming Tests) (For All Data)
DAF model 0.999938295 0.998425175 0.999508986
FF model 0.999658564 0.998359221 0.999253618
113
Table 39: The amount of experimental and predicted data and their errors for the confirmation and optimized
experiments.
Error of Error of
Sample’s Experimental Predicted Experimental Predicted
Model for Model for
Code DAF (N) DAF (N) FF (N) FF (N)
DAF FF
CT1 3670 3851 4.9 % 6812 7008 2.9 %
CT2 3920 4060 3.6 % 7066 6738 4.6 %
CT3 3280 3151 3.9 % 7660 7647 0.2 %
CT4 4220 4028 4.6 % 7653 7291 4.7 %
Opt1 3650 3730 2.2 % 7780 7392 5.0 %
|𝐴𝑖 −𝑌𝑖 |
(Error of model = × 100)
𝐴𝑖
6.5.4 Effect of process parameters on the downward axial force and failure force
Figure 83 demonstrates the correlation between the tool plunge depth, traverse speed and rotational
speed with the downward axial force. The DAF increases with the increase in the tool traverse speed
due to the colder material in front of the tool at higher traverse speeds. Therefore, more force is needed
to pass the tool through the cold material of the workpiece [1]. As seen in Figure 83, the increase in
tool rotational speed causes the reduction of the DAF. It can be attributed to the increase in the heat
input that softened the material ahead of the tool during welding [1]. The increase of the tool plunge
depth causes the increase of DAF because the frictional contact between the tool and the workpiece
material increases at high plunge depths [14, 109]. Figure 84 illustrates the relationship between the
tool plunge depth, traverse speed, rotational speed and the failure force. It is obvious that the increase
of tool traverse speed leads to the reduction of FF because it affects the mixing and stirring processes,
negatively, and causes the formation of defects and a weak joint [56, 110]. The FF reduces with the
increase in tool rotational speed due to additional heat input generated at high rotational speeds. It
also causes the widening of the HAZ area that affects the strength of the joint, adversely [1]. At first,
the increase in tool plunge depth leads to the increase in the FF probably due to stronger mixing and
stirring of the material. However, at higher plunge depths, the FF decreases possibly due to the
formation of more severe hooking defects and the upper sheet thinning in the welded area [56, 102,
111]. Having a joint with a high failure force while keeping the downward axial force minimized is
promising. Because the higher failure force, the higher the quality of the joint. On the other hand, the
lower downward axial force during welding, the better because one can use lower price CNC
machines with lower capacity and stiffness. Moreover, there is no need for a lot of clamping and
fixturing at low downward axial forces. Therefore, making a joint at lower axial forces helps to reduce
the equipment costs. Thus, a compromise between the fracture force and the downward axial force
yields the best condition to make the welds by RAFSW technique. According to Figure 83 and Figure
84, the efficient working window of the process parameters is shown in Table 40. As seen, the
114
optimized tool traverse speed could be much higher than the presented ones in the previous paper
[115]. Higher traverse speeds make the process more desirable for industrial use. Based on the
developed ANN models, an optimized condition has been tested and reported in Table 34. In this
condition, the traverse speed is as high as 1700 mm/min. Thus, a compromise between the fracture
force and the downward axial force yields the best condition to make the welds by RAFSW technique.
Figure 83: Contour plots of the DAF versus: (a) traverse speed and rotational speed (when the plunge depth is
2.65 mm), (b) plunge depth and rotational speed (when the traverse speed is 1700 mm/min), (c) plunge depth
and traverse speed (when the rotational speed is 4400 rpm). The contour plots are extracted from developed
ANN models in this study.
Figure 84: Contour plots of the FF versus: (a) traverse speed and rotational speed (when the plunge depth is
2.65 mm), (b) plunge depth and rotational speed (when the traverse speed is 1700 mm/min), (c) plunge depth
and traverse speed (when the rotational speed is 4400 rpm). The contour plots are extracted from developed
ANN models in this study.
Table 40: An efficient working window of the process parameters.
V (mm/min) W (rpm) PD (mm) OL (mm)

1400–1800 3400-4500 2.575–2.675 11.5-19
115
6.6 Adaptation of the RAFSW process
6.6.1 Adaptation of RAFSW technique and its apparatus
To adapt the RAFSW technique to industrial applications, it is necessary to adjust it so that it tolerates
more joint fit-up variations. Moreover, it should have low sensitivity to positioning and aligning and
do not need lots of clamping and stiff CNC machines.
FSW can be done in either position control or force control mode. In position control the plunge depth
is kept constant through the weld path. It is difficult to acquire a quality weld in position control mode
in presence of positioning inaccuracies or part dimensional inconsistency or when the machine is not
rigid enough [4]. FSW in force control can tolerate higher levels of joint fit-up disturbances and
inaccuracy in positioning and aligning. Moreover, force control mode can correct the positioning
errors and dimensional inconsistencies to yield a flawless weld [116]. The development of the
RAFSW technique in force control could lead to a wider range of applications. Moreover, it would
be possible to use big-size, cheap CNC machines with low stiffness and accuracy. Besides, there
would be no need for sturdy fixtures and clamping as the spring-loaded tool holder squeezes the parts
together like a mobile fixture through the weld path.
Unless expensive specialized FSW machines are used, typical CNC machining centers or large
routers can operate only in position control. To alleviate this problem, a spring loaded tool holder
has been developed and used in this section as shown in Figure 85. As set for the following
experiments, the tool holder allows a variation of only 500 N over an axial stroke of +/- 0.5 mm
relative to its set point. The set point is centered at 3650 N as observed in Table 34. By applying this
set point in the welding program, a nearly constant force of 3650 N is obtained assuming dimensional
variations of less than 0.5 mm about the set point [4].
In force control, a tool may excessively plunge into the workpiece [18, 117]. To address this issue,
the previous tool design has been updated for force control mode. As shown in Figure 85.a, the new
tool design includes tapered entry and rounded edges. The depth and width of the tapered entry feature
are 0.25, and 3.5 mm, respectively. Using this design, a slight increase in the plunge depth causes a
big increase in the shoulder area in contact with the workpiece. This means that a high force is needed
for a slight increase in the tool plunge depth. Therefore, the tapered entry avoids excessive plunging
of the tool into the workpiece and cutting the workpiece material instead of mixing and stirring.
Moreover, it prevents big fluctuations of the force during force control welding.
116
6.6.2 Validation of the tool and tool holder
The developed RAFSW technique in force control using a spring-loaded tool holder and tapered entry
tool is expected to be capable of compensating high amounts of workpiece dimensional
inconsistencies, positioning and aligning inaccuracies, and be applicable with light clamping by low
CNC machine stiffness. To verify these characteristics, some welds were conducted by the RAFSW
technique in force control. At first, a pair of overlapped metal sheets were mounted on the Fryer
MC15 CNC machine. The sheets were completely clamped like what we had in position control
welding. The welding conditions were the same as optimized conditions presented in Table 34. The
plunge height was set according to the set point of the spring-loaded tool holder. During welding,
the amount of downward axial force was approximately equal to the force measured in position
control. Tensile test samples also demonstrated a failure force close to the failure force of the samples
welded in position control.
In the next step, a small delta-shaped assembly was mounted on the Fryer MC15 CNC machine. This
sample had a lot of part dimensional inconsistency and positioning inaccuracies. As shown in Figure
86.a, the sample was fixed just at two corners on one side and was free on the other side so that there
is a big gap between the sheets on the free side. As can be seen in Figure 86.b and c, the weld was
done on both sides, correctly. Hence, the developed technique could compensate for all the
inconsistencies, inaccuracies, and lack of sturdy fixturing.
Figure 85: (a) The new tool with the tapered entry, (b) the spring-loaded tool holder, (c) the whole set-up for
the RAFSW in force control on Fryer MC15 CNC machine, and (d) a welded sample at the optimized welding
condition.
117
Figure 86: Welding process of a delta-shaped structure clamped partially on one side on the MC15 machine.
The part is installed on a dynamometer plate: (a) The clamped structure on one side. There is a considerable
gap between the sheets in the other side, (b) One side is welded. The other side is under welding (the gap is
compensated; and the sheets are squeezed together), (c) The welded sample in both sides.
6.7 Application and test of larger panel assemblies

6.7.1 CNC router set-up and welding sequence
The same RAFSW technique has been implemented on a big CNC router that was less stiff than the
Fryer MC15 machine. Larger delta-shaped assemblies were clamped just in the corners, as shown in
Figure 87.a. Before the main welding process, it could be useful to weld small tags on the structure.
They can work as fixtures to keep the sheets together and may decrease the distortion of the assembled
structure. Thus, the effect of the weld tags on the distortion of the assembled structure was explored.
As illustrated in Figure 87.c, the first sample was fixed with clamps at the corners without weld tags.
The second and third samples were fixed identical to the first sample and had 10 cm long weld tags
prior to continuous welding. The assembled samples are shown in Figure 87.a and b. The location of
the fixtures and tags are shown in Figure 87.c. The distortion profile of the samples is illustrated in
Figure 87.d. It can be concluded that the second sample shows not only a more even distortion
distribution but also the minimum distortion level. Figure 87.e illustrates a schematic of the welding
procedure. The location of advanced side (AS) and retreating side (RS) are indicated for each weld
path. Figure 87.f shows the macrostructure of both welded sides of the panel. From Figure 87.e and
f, it is clear that the welded sample has a symmetry in terms of the location of the AS and RS for each
weld path. It means that the RS is always near the center of the welded side panel and the AS is always
located toward the edge of the side panel. This kind of configuration leads to symmetrical behavior
of the entire structure despite the asymmetrical nature of one-path FSW joints.
118
(e)
(f)
(AS) (RS)
(RS) (AS)
Figure 87: (a) The RAFSW process in force control mode to weld a large truck side panel on a big-size CNC
router, (b) the welded samples at 3 different welding sequences with and without weld tags, (c) the top view
of the clamped and tagged panels (the location of tags and clamps are shown in yellow and orange,
respectively), (d) distortion distribution of each sample (e) schematic of the procedure with the identification
of the advanced side (AS) and retreating side (RS) of each weld path, (f) the macrostructure of welded
samples.
6.7.2 Strength validation of truck side panels

For the need of this paper, the strength of truck panel samples has been validated using finite element
simulations followed by experiments. The simulations are used to benchmark the level of load that
can be sustained while the experimental tests identify the real strength. Two load cases are presented
next: (1) a traction load normal to the face of the stiffener applied on 76.2-mm-long coupons taken
from longer samples of the section 5-1 of the paper, (2) cyclic loads applied in compression on 304.8-
mm-long coupons.
119
6.7.2.1. Traction test
For FE simulations, NX-NASTRAN software is used to design the assembled coupon and clamps on
a hydraulic press equipped with a load cell as shown in Figure 88. The purpose of the simulation is
to establish the location of the stress concentration. It shows that the maximum stress is concentrated
in the HAZ area of the weld and near the lower bent part of the delta-shaped sheet. The stress in the
lower bent part of the delta-shaped sheet and the crack-like region of the weld are about 135 and 85
MPa, respectively. In the HAZ area, the stress is maximum in the crack-like region of the weld
between the two welded sheets. Hook and cold lap defects are potential failure sites located in this
place, too [58, 101, 102, 118, 119]. The HAZ has lower mechanical properties and yields at lower
stresses compared to the parent material [56]. It can be seen that the applied tensile load transfers to
the joint as if one tries to disassemble the joint. The stresses are high in the crack-like area in the joint.
Therefore, in the designed test, high stresses in the welded area can cause the failure of the joint.
Although the stresses are relatively higher in the lower bent area than the weld area, the weld still has
a high chance to be the failure site. Because the bending of the sheet leads to work hardening of the
bent area which boosts the mechanical properties of this region compared to the parent material, while
the weld area has lower mechanical properties than the parent material.
The same configuration has been tested as illustrated in Figure 89. The 76.2-mm-long welded coupon
failed at 8885.72 N of traction force. The failure happens somewhere near the bent part of the delta-
shaped sheet as can be seen in Figure 89.b. A close look at the welded region reveals that the joint
started to fail from the crack-like area, too. Figure 89.c shows that a crack initiated from the crack-
like region of the weld. Therefore, one can say that the failure has started in both bent region of the
delta-shaped and the weld area, but the breakage happens in the bent area sooner. Briefly, the static
force sustained by a coupon of this size is high enough for merchandize anchored to a full-size
assembled side panel of a truck box.
Applied load
Figure 88: The stress distribution (Von-misses criterion) for a 3-inch-long weld coupon under a 4448.22 N
tensile load.
120
Figure 89: (a) the experimental set-up to test weld coupon under static load in tension, (b) the broken sample
after testing, (c) a close look to the broken area and the crack initiation site in the weld area.
6.7.2.2. Fatigue tests in compression
The next step is to study the effect of cyclic loads. The design presented in Figure 90 shows the stress
distribution in the sample simulated with 1200 N and 2880 N compression loads. The compressive
load on top of the structure results in tensile stresses in the welded area. The simulations show that
the structure deflects elastically at both load cases. As can be seen in Figure 90, the maximum stress
concentration is in the red regions on the bent area of the delta-shaped part. However, the stress is in
compression which does not cause failure of the upper sheet under cyclic loads. The maximum tensile
stress was found in the crack-like area of the weld region. Indeed, the HAZ area will yield sooner
than the parent material at a given stress since it has a lower hardness and yield strength than the base
material and weld nugget zone [56, 119]. Moreover, sharp crack-like edges in the weld region act as
potential places for crack initiation during cyclic loading [119]. Thus, the first expected place to yield
in tension is the crack-like region between the sheets in the welded area. As can be seen in Figure
90.b, the stress is about 89 MPa in the corners of the welded area while it is 115 MPa in the middle
of the weld seam under a 2880 N load. Hence, it is expected that the fatigue crack starts from the
middle of the weld seam in the crack-like area and propagates toward the corners of the joint along
the weld path. Consequently, it is presumed that the designed set-up could evaluate the strength of
the joint under the intended loading. Fatigue tests in compression has been carried out using 304.8
121
mm-long coupons, illustrated in Figure 91. The fatigue tests were done at a frequency of 2 Hz and a
R ratio of approximatively 0.05 at various compressive loads. The sample has not failed under a
maximum force of 1200 N during the first million cycles. It failed after about 100,000 cycles later as
the maximum applied load was increased to 2880 N. Figure 92 shows the failed sample from different
views. As shown in Figure 92, the failure started from the center of the weld seam as predicted by the
simulations. As expected, the fracture was happened along the weld seam. Figure 92.c shows a close
view of the joint of the failed part. From this figure, it can be concluded that the crack-like region in
the weld area was the fatigue crack initiation site that is in conjunction with the simulation results.
Generally, both set-ups worked as expected to evaluate the strength of the assembled parts. The results
show that the assembled truck side panel is capable of standing the studied static and cyclic load
cases. Thus, the developed RAFSW technique at force control mode on a low-cost CNC router is a
reliable, cost-effective technique for industrial applications such as assembly of side panels of truck
box.
Figure 90: The stress distribution for a 1-ft-long weld coupon under a (a) 1200 N and (b) 2880 N compressive
load. The magnified, cross-sectioned view of the joint is shown both in the corner and middle of the
assembled parts. The amount of tensile stresses in the the crack-like area of the joint are identified in these
cross-sections.
122
Figure 91: the experimental set-up to test weld coupon under dynamic load in compression.
Figure 92: (a) the broken sample after fatigue test, (b) as can be seen, maximum plastic deformation is located
in the middle of the weld seam; the crack initiated from the middle of weld area, (c) a close look to the broken
corner demonstrates that the crack initiated from the crack-like region of the weld.
6.8 Conclusion
Based on the results obtained in this research, the following conclusions are derived:
• An efficient working window of the RAFSW process parameters is found at traverse speeds
as high as 1700 mm/min for lap joint configuration.
• The RAFSW technique has been successfully applied in force control mode using a spring-
loaded tool holder and a new tool design with a tapered entry.
123
• The developed technique could not only work on low-cost CNC machines but also on large,
low-cost CNC routers with less stiffness than the needed machines for RAFSW in position
control.
• This technique is capable of bearing high levels of positioning and aligning inaccuracies, part
dimensional inconsistencies, lack of machine rigidity and rigid clamping. As the developed
force control mode equipped with the tapered entry tool, it could prevent the tool from
digging into the workpiece material while squeezing the two sheets together.
• The developed technique has been successfully implemented on a big low-cost, low-stiffness
CNC router. Large truck side panels were assembled by the technique with minimized
clamping. The structures were just fixed at the corners.
• The distortion of the large assembled parts is minimized using weld tags in the middle of the
weld path prior to welding.
• The static test with a tensile load on a 76.2-mm sample showed that the sample could bear
8885 N. The failure starts from two locations located in the lower bent area of the upper sheet
and the crack-like area of the weld.
• The fatigue test shows that a 1-ft-long assembled coupon does not fail under a 1200 N force
up to one million cycle. It fails after 100,000 cycles under 2880 N load. The failure happens
along the joint area. The crack initiated from the center of the weld seam.
In summary, this study establishes the working guidelines for the RAFSW technique developed in
force control mode on a low-cost CNC router for industrial applications, specifically to assemble
large side panels of truck box.
Funding: This research has been supported by funds of PI2 Team.
Acknowledgments: The authors would like to thank the PI2/REGAL team members who made this
research possible.
124
Conclusions and Perspectives
Overview of the project
FSW is a solid-state metal joining process with many advantages over fusion welding techniques.
However, machine costs and royalties act as the main barriers against its widespread use. To make
the process cost-effective, PI2/REGAL team at Laval University has recently developed a low-cost
FSW technique at the right angle (RAFSW). To provide an overview, there are five or six FSW
machines available in the Province of Quebec, while there are many thousands of CNC milling
machines capable of operating the proposed RAFSW technique. The main goal of this thesis was to
continue to develop the different aspects of the RAFSW technique to make it reliable and affordable
for potential industrial users. To this end, at first, a full study was done on the effect of RAFSW
process parameters on generated forces and mechanical properties of butt joints welded at high
traverse speeds. Afterward, an investigation was conducted on the effect of post weld heat treatment
to recover the mechanical properties of the optimized welds. Moreover, the technique has been
developed for lap joint configuration. Finally, the RAFSW technique was adapted to work in the
force control to assemble large truck side panels.
Conclusions
1- A research was conducted to gain a deep insight through the effect of RAFSW process parameters
on generated axial forces and tensile properties of AA6061-T6 butt-welded plates of 6.35 mm thick.
The effect of process parameters on generated force and UTS of the joint was studied using Taguchi
method and ANN modeling. The efficient working window of process parameters is obtained when
the UTS is high while the downward axial force is low. One can make quality joint with a UTS more
than 247 MPa at traverse speeds as high as 600 mm/min. Moreover, the robustness of the process was
demonstrated in the presence of joint fit-up disturbances. It was shown that a standard artificial aging
process could recover the mechanical properties of the weld, partially. Finally, the microstructure and
micro-hardness of the joint made at optimized condition were evaluated. The results of this part of
the thesis demonstrate the reliability of the technique to make butt-joints at high traverse speeds.
2- The aim of this part was to analyze the effect of various post weld heat treatments (PWHT) on the
mechanical and physical properties of the RAFSW joints. To this end, the optimized process
parameters from the previous part of the research were used to weld butt joints of AA6061-T6 alloy.
A comprehensive study was done on the effect of natural aging, artificial aging, and T6 heat
treatments on the mechanical and physical properties of welded samples. The optimized conditions
125
for artificial aging were obtained. It was found that the joint efficiency of RAFSW samples reaches
90% by artificial aging at 180 °C for 18 h. Moreover, it was revealed that industrial users can take
advantage of a fast artificial aging process at 220 °C for 30 min to obtain a RAFSW joint with 85%
of joint efficiency and 241 MPa of tensile strength. The repeatability of the PWHTs was demonstrated
when there is a delay to do the artificial aging process. It was shown that solubilizing heat treatment
before artificial aging at 200°C for 2 h results in improvement of the tensile strength up to 243 MPa
while the ductility decreases compared to the samples just aged artificially. It can be concluded that
a single fast artificial aging process is more beneficial than solubilizing followed by an artificial aging
process from an industrial perspective.
3- The main goal of the next part of the thesis was to evaluate the effect of the tool design and process
parameters on the downward axial force and the failure force of the lap joints made by the RAFSW
technique. The efficient range was obtained for tool design and process parameters to make quality
lap joints at good traverse speeds using the Taguchi method and ANN modeling. It was identified
that one can achieve a quality lap joint at traverse speeds as high as 1400 mm/min and downward
axial forces as low as 3.2 kN. Thus, optimized tool design and its efficient working window of process
parameters were obtained to make lap joints by the RAFSW technique.
4- In this part of the thesis, the goal was to develop the RAFSW technique to assemble large side
panels of the truck box so that the technique be industrially reliable and cost-effective. At first, the
efficient working window of the welding process parameters was updated at higher traverse speeds
using Taguchi method and ANN modeling for the optimized tool design from the previous part of the
thesis. The results showed that quality lap joints can be made at traverse speeds as high as 1700
mm/min. Afterward, the technique was developed to operate in force control mode using a spring
load tool holder and a newly designed tool with a tapered entry. This technique is capable of bearing
high levels of positioning and aligning inaccuracies, part dimensional inconsistencies, lack of
machine rigidity and rigid clamping. It was shown that the technique could be operated on big, low
stiffness CNC routers with minimized clamping. Large truck side panels were welded by the
technique while the distortion is kept low. Some set-ups were designed to study the effect of static
and dynamic loading on the assembled truck panels using NX-NASTRAN software. The fatigue test
showed that the assembled truck side panel does not fail under a 1200 N force up to one million
cycles. It fails after 100,000 cycles under 2880 N load.
126
Perspectives
Based on the results obtained from this work, there is still room for more improvement. The following
recommendations are presented for future work:
• Bring the tool design to a commercial level of reliability with all additional features and
evaluating the wear resistance of the tools
• Develop the technique for T-joints
• Study the effect of fatigue for different configurations and structures
• Study the joining of other alloys
• Study thicknesses from 1/16 inch to ½ inch
• Applying the techniques to different structures for industrial use
• Design of fixtures and clamping
• Minimizing the distortion of the large assembled structures
• Design of hybrid joints
• Design of extrusions for the RAFSW process
• Welding of dissimilar alloys
127
References
[1] R.S. Mishra, Z.Y. Ma, Friction stir welding and processing, Materials Science & Engineering R
50(1) (2005) 1-78.
[2] R. Nandan, T. Debroy, H.K.D.H. Bhadeshia, Recent advances in friction-stir welding – Process,
weldment structure and properties, Progress in Materials Science 53(6) (2008) 980-1023.
[3] G. Mathers, The Welding of Aluminium and its Alloys, Woodhead Publishing Ltd, Abington Hall,
Abington, Cambridge CB1 6AH, England, 2002.
[4] B.T. Gibson, D.H. Lammlein, T.J. Prater, W.R. Longhurst, C.D. Cox, M.C. Ballun, K.J.
Dharmaraj, G.E. Cook, A.M. Strauss, Friction stir welding: Process, automation, and control, Journal
of Manufacturing Processes 16(1) (2014) 56-73.
[5] A.P. Reynolds, Microstructure Development in Aluminum Alloy Friction Stir Welds, in: R.S.
Mishra, M.W. Mahoney (Eds.), Friction stir welding and processing, ASM International, Materials
Park, Ohio, 2007, pp. 51-70.
[6] S. Kou, Welding metallurgy, Wiley, New York, 1987.
[7] M.W. Mahoney, Mechanical Properties of Friction Stir Welded Aluminum Alloys, in: R.S.
Park, Ohio, 2007, pp. 71-110.
[8] W.J. Arbegast, Application of Friction Stir Welding and Related Technologies, in: R.S. Mishra,
M.W. Mahoney (Eds.), Friction stir welding and processing, ASM International, Materials Park,
Ohio, 2007, pp. 273-308.
[9] C. Liu, X. Yi, Residual stress measurement on AA6061-T6 aluminum alloy friction stir butt welds
using contour method, Materials and Design 46 (2013) 366-371.
[10] S. Gachi, F. Boubenider, F. Belahcene, Residual stress, microstructure and microhardness
measurements in AA7075-T6 FSW welded sheets, Nondestructive Testing and Evaluation 26(1)
(2011) 1-11.
[11] J. He, Z. Ling, H. Li, Effect of tool rotational speed on residual stress, microstructure, and tensile
properties of friction stir welded 6061-T6 aluminum alloy thick plate, The International Journal of
Advanced Manufacturing Technology 84(9) (2016) 1953-1961.
[12] S. Rajakumar, V. Balasubramanian, Establishing relationships between mechanical properties of
aluminium alloys and optimised friction stir welding process parameters, Materials and Design 40
(2012) 17-35.
[13] C.B. Fuller, Friction Stir Tooling: Tool Materials and Designs, in: R.S. Mishra, M.W. Mahoney
(Eds.), Friction stir welding and processing, ASM International, Materials Park, Ohio, 2007, pp. 7-
35.
[14] R. Rai, A. De, H.K.D.H. Bhadeshia, T. Debroy, Review: friction stir welding tools, Science and
Technology of Welding and Joining 16(4) (2011) 325-342.
[15] N. Mendes, P. Neto, A. Loureiro, A.P. Moreira, Machines and control systems for friction stir
welding: A review, Materials & Design 90 (2016) 256-265.
[16] https://www.fooke-portalfraesmaschinen.de/.
[17] https://www.bondtechnologies.net/.
[18] C.B. Smith, Robots and Machines for Friction Stir Welding/Processing, in: R.S. Mishra, M.W.
Mahoney (Eds.), Friction stir welding and processing, ASM International, Materials Park, Ohio,
2007, pp. 219-233.
[19] R.S. Mishra, M.W. Mahoney, Future Outlook for Friction Stir Welding and Processing, in: R.S.
Park, Ohio, 2007, pp. 351-352.
[20] Y. Imani, M. Guillot, Axial Force Reduction in Friction Stir Welding of AA6061-T6 at Right
Angle, ASME 2014 International Mechanical Engineering Congress and Exposition, American
Society of Mechanical Engineers, Montreal, Canada, 2014, pp. V02BT02A016-V02BT02A016.
128
[21] P.H. Shah, V.J. Badheka, An Experimental Insight on the Selection of the Tool Tilt Angle for
Friction Stir Welding of 7075 T651 Aluminum Alloys, Indian Journal of Science and Technology
9(S1) (2017).
[22] E.F. Shultz, E.G. Cole, C.B. Smith, M.R. Zinn, N.J. Ferrier, F.E. Pfefferkorn, Effect of
Compliance and Travel Angle on Friction Stir Welding With Gaps, Journal of Manufacturing Science
and Engineering 132(4) (2010) 041010.
[23] H. Takahara, M. Tsujikawa, S.W. Chung, Y. Okawa, K. Higashi, S. Oki, Optimization of
Welding Condition for Nonlinear Friction Stir Welding, MATERIALS TRANSACTIONS 49(6)
(2008) 1359-1364.
[24] Y. Imani, M. Guillot, A. Tremblay, Study and Application of Complex 3D Friction Stir Welding
of Butt and Lap Joints Using a 5-axis CNC Mashine Tool, CSME Proceedings of the Canadian
Society for Mechanical Engineering International Congress, CSME, Toronto, Canada, 2014.
[25] K. Mehta, V. Badheka, Influence of tool design and process parameters on dissimilar friction stir
welding of copper to AA6061-T651 joints, The International Journal of Advanced Manufacturing
Technology 80(9) (2015) 2073-2082.
[26] M. Melendez, W. Tang, C. Schmidt, J.C. McClure, A.C. Nunes, L.E. Murr, Tool Forces
Developed During Friction Stir Welding, NASA Technical Reports Server
nasa_techdoc_20030071631 (2003).
[27] Y. Imani, M. Guillot, A. Tremblay, Optimization of FSW Tool Design and Operating Parameters
for Butt Welding of AA6061-T6 at Right angle, CSME Proceedings of the Canadian Society for
Mechanical Engineering International Congress, Toronto, Canada, 2014.
[28] Y. Imani, M. Guillot, Tolerating for joint fit-up issues using welding parameters in friction stir
welding of AA6061-T6 at right angle, CSME Proceedings of the Canadian Society for Mechanical
Engineering International Congress, CSME, Toronto, Canada, 2014.
[29] Y. Imani, Development of friction stir welding techniques for multi-axis machines, Mechanical
Engineering, Laval University, Canada, Quebec, 2015.
[30] O. Marcotte, développement d’un système de contrôle pour la robotisation du soudage par
friction malaxage, Mechanical Engineering, Laval University, Canada, Quebec, 2009.
[31] D. Trimble, G.E. O’donnell, J. Monaghan, Characterisation of tool shape and rotational speed
for increased speed during friction stir welding of AA2024-T3, Journal of Manufacturing Processes
17 (2015) 141-150.
[32] K. Aota, H. Okamura, K. Sato, Friction stir welding method for reducing the friction force, in:
E.P. Office (Ed.) European Patent Office, Hitachi, Ltd. Chiyoda-ku, Tokyo (JP), 2003.
[33] R. Kumar, K. Singh, S. Pandey, Process forces and heat input as function of process parameters
in AA5083 friction stir welds, Transactions of Nonferrous Metals Society of China 22(2) (2012) 288-
298.
[34] K. Kumar, S.V. Kailas, On the role of axial load and the effect of interface position on the tensile
strength of a friction stir welded aluminium alloy, Materials and Design 29(4) (2008) 791-797.
[35] W. Safeen, S. Hussain, A. Wasim, M. Jahanzaib, H. Aziz, H. Abdalla, Predicting the tensile
strength, impact toughness, and hardness of friction stir-welded AA6061-T6 using response surface
methodology, The International Journal of Advanced Manufacturing Technology 87(5) (2016) 1765-
1781.
[36] A. Heidarzadeh, H. Khodaverdizadeh, A. Mahmoudi, E. Nazari, Tensile behavior of friction stir
welded AA 6061-T4 aluminum alloy joints, Materials and Design 37 (2012) 166-173.
[37] S. Rajakumar, C. Muralidharan, V. Balasubramanian, Predicting tensile strength, hardness and
corrosion rate of friction stir welded AA6061-T 6 aluminium alloy joints, Materials and Design 32(5)
(2011) 2878-2890.
[38] G. Elatharasan, V.S.S. Kumar, An Experimental Analysis and Optimization of Process
Parameter on Friction Stir Welding of AA 6061-T6 Aluminum Alloy using RSM, Procedia
Engineering 64 (2013) 1227-1234.
129
[39] P. Ramulu, R. Narayanan, S. Kailas, J. Reddy, Internal defect and process parameter analysis
during friction stir welding of Al 6061 sheets, The International Journal of Advanced Manufacturing
Technology 65(9) (2013) 1515-1528.
[40] J.D. Costa, J.A.M. Ferreira, L.P. Borrego, Influence of spectrum loading on fatigue resistance of
AA6082 friction stir welds, International Journal of Structural Integrity 2(2) (2011) 122-134.
[41] J.D. Costa, J.A.M. Ferreira, L.P. Borrego, L.P. Abreu, Fatigue behaviour of AA6082 friction stir
welds under variable loadings, International Journal of Fatigue (2011).
[42] C. He, Y. Liu, J. Dong, Q. Wang, D. Wagner, C. Bathias, Fatigue crack initiation behaviors
throughout friction stir welded joints in AA7075-T6 in ultrasonic fatigue, International Journal of
Fatigue 81 (2015) 171-178.
[43] M. Ericsson, R. Sandström, Influence of welding speed on the fatigue of friction stir welds, and
comparison with MIG and TIG, International Journal of Fatigue 25(12) (2003) 1379-1387.
[44] S. Lomolino, R. Tovo, J. Dos Santos, On the fatigue behaviour and design curves of friction stir
butt-welded Al alloys, International Journal of Fatigue 27(3) (2005) 305-316.
[45] P. Wanjara, B. Monsarrat, S. Larose, Gap tolerance allowance and robotic operational window
for friction stir butt welding of AA6061, Journal of Materials Processing Tech. (2012).
[46] E. Cole, A. Fehrenbacher, E. Shultz, C. Smith, N. Ferrier, M. Zinn, F. Pfefferkorn, Stability of
the friction stir welding process in presence of workpiece mating variations, The International Journal
of Advanced Manufacturing Technology 63(5) (2012) 583-593.
[47] E.F. Shultz, A. Fehrenbacher, F.E. Pfefferkorn, M.R. Zinn, N.J. Ferrier, Shared control of robotic
friction stir welding in the presence of imperfect joint fit-up, Journal of Manufacturing Processes
(2012).
[48] Z. Zhang, B.L. Xiao, Z.Y. Ma, Enhancing mechanical properties of friction stir welded 2219Al-
T6 joints at high welding speed through water cooling and post-welding artificial ageing, Materials
Characterization 106 (2015) 255-265.
[49] G. Ipekoglu, S. Erim, G. Cam, Investigation into the Influence of Post-Weld Heat Treatment on
the Friction Stir Welded AA6061 Al-Alloy Plates with Different Temper Conditions, Metallurgical
and Materials Transactions A 45A(2) (2014) 864-877.
[50] E.A. El-Danaf, M.M. El-Rayes, Microstructure and mechanical properties of friction stir welded
6082 AA in as welded and post weld heat treated conditions, Materials and Design 46 (2013) 561-
572.
[51] G. Ipekoglu, S. Erim, G. Cam, Effects of temper condition and post weld heat treatment on the
microstructure and mechanical properties of friction stir butt-welded AA7075 Al alloy plates,
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY 70(1-4)
(2014) 201-213.
[52] P. Sivaraj, D. Kanagarajan, V. Balasubramanian, Effect of post weld heat treatment on tensile
properties and microstructure characteristics of friction stir welded armour grade AA7075-T651
aluminium alloy, Defence Technology 10(1) (2014) 1-8.
[53] Y. Zhao, L. Zhou, Q. Wang, K. Yan, J. Zou, Defects and tensile properties of 6013 aluminum
alloy T-joints by friction stir welding, Materials and Design 57 (2014) 146-155.
[54] A.C.F. Silva, D.F.O. Braga, M.A.V.d. Figueiredo, P.M.G.P. Moreira, Friction stir welded T-
joints optimization, Materials and Design 55 (2014) 120-127.
[55] L. Cui, X. Yang, Y. Xie, X. Hou, Y. Song, Process parameter influence on defects and tensile
properties of friction stir welded T-joints on AA6061-T4 sheets, Materials and Design 51 (2013) 161-
174.
[56] H. Liu, Y. Zhao, Y. Hu, S. Chen, Z. Lin, Microstructural characteristics and mechanical
properties of friction stir lap welding joint of Alclad 7B04-T74 aluminum alloy, The International
Journal of Advanced Manufacturing Technology 78(9) (2015) 1415-1425.
[57] E. Salari, M. Jahazi, A. Khodabandeh, H. Ghasemi-Nanesa, Influence of tool geometry and
rotational speed on mechanical properties and defect formation in friction stir lap welded 5456
aluminum alloy sheets, Materials and Design 58 (2014) 381-389.
130
[58] G. D'Urso, C. Giardini, The influence of process parameters and tool geometry on mechanical
properties of friction stir welded aluminum lap joints, International Journal of Material Forming 3(1)
(2010) 1011-1014.
[59] X. Hou, X. Yang, L. Cui, G. Zhou, Influences of joint geometry on defects and mechanical
properties of friction stir welded AA6061-T4 T-joints, Materials and Design 53 (2014) 106-117.
[60] G. Buffa, G. Campanile, L. Fratini, A. Prisco, Friction stir welding of lap joints: Influence of
process parameters on the metallurgical and mechanical properties, Materials Science & Engineering
A 519(1) (2009) 19-26.
[61] M. Mahdavi Shahri, R. Sandström, W. Osikowicz, Critical distance method to estimate the
fatigue life time of friction stir welded profiles, International Journal of Fatigue (2011).
[62] M.M. Shahri, R. Sandström, Fatigue analysis of friction stir welded aluminium profile using
critical distance, International Journal of Fatigue 32(2) (2010) 302-309.
[63] L.G. Vigh, I. Okura, Fatigue behaviour of Friction Stir Welded aluminium bridge deck segment,
Materials and Design 44 (2013) 119-127.
[64] G.K. Padhy, C.S. Wu, S. Gao, Friction stir based welding and processing technologies -
processes, parameters, microstructures and applications: A review, Journal of Materials Science &
Technology (2017).
[65] E.S. Abdel-All, M.C. Frank, I.V. Rivero, Rapid tooling using friction stir welding and
machining, Rapid Prototyping Journal 23(1) (2017) 81-95.
[66] J.J.S. Dilip, S. Babu, S.V. Rajan, K.H. Rafi, G.D.J. Ram, B.E. Stucker, Use of Friction Surfacing
for Additive Manufacturing, Materials and Manufacturing Processes 28(2) (2013) 189-194.
[67] C. Shah, B. Goyal, V. Patel, Optimization of FSW Process Parameters for AlSiCp PRMMC
Using ANOVA, Materials Today: Proceedings 2(4-5) (2015) 2504-2511.
[68] K. Mehta, V. Badheka, Experimental Investigation of Process Parameters on Defects Generation
in Copper to AA6061-T651 friction stir Welding, International Journal of Advances in Mechanical
and Automobile Engineering 3(1) (2016) 4.
[69] S. Rajakumar, C. Muralidharan, V. Balasubramanian, Establishing empirical relationships to
predict grain size and tensile strength of friction stir welded AA 6061-T6 aluminium alloy joints,
Transactions of Nonferrous Metals Society of China 20(10) (2010) 1863-1872.
[70] M. Momeni, M. Guillot, Development of friction stir welding technique at right angle (RAFSW)
applied on butt joint of AA6061-T6 aluminum alloy, International Journal of Advanced
Manufacturing Technology 99(9) (2018) 3077-3089.
[71] M. Pishevar, J. Mohandesi, H. Omidvar, M. Safarkhanian, Influences of Friction Stir Welding
Parameters on Microstructural and Mechanical Properties of AA5456 (AlMg5) at Different Lap Joint
Thicknesses, Journal of Materials Engineering and Performance 24(10) (2015) 3835-3844.
[72] R.V. Rao, SpringerLink, Advanced modeling and optimization of manufacturing processes
international research and development, Springer, London, 2011.
[73] B. Das, S. Pal, S. Bag, Torque based defect detection and weld quality modelling in friction stir
welding process, Journal of Manufacturing Processes 27 (2017) 8-17.
[74] N.D. Ghetiya, K.M. Patel, Prediction of Tensile Strength in Friction Stir Welded Aluminium
Alloy Using Artificial Neural Network, Procedia Technology 14 (2014) 274-281.
[75] G. Buffa, L. Fratini, F. Micari, Mechanical and microstructural properties prediction by artificial
neural networks in FSW processes of dual phase titanium alloys, Journal of Manufacturing Processes
14(3) (2012) 289-296.
[76] H. Okuyucu, A. Kurt, E. Arcaklioglu, Artificial neural network application to the friction stir
welding of aluminum plates, Materials and Design 28(1) (2007) 78-84.
[77] R. Palanivel, R. Laubscher, I. Dinaharan, N. Murugan, Tensile strength prediction of dissimilar
friction stir-welded AA6351–AA5083 using artificial neural network technique, Journal of the
Brazilian Society of Mechanical Sciences and Engineering 38(6) (2016) 1647-1657.
[78] L. Gao, Y. Harada, S. Kumai, Microstructural characterization of aluminum alloys using Weck's
reagent, part I: Applications, Materials Characterization 107 (2015) 426-433.
131
[79] W. Tang, X. Guo, J. McClure, L. Murr, Heat input and temperature distribution in friction stir
welding, Journal of Materials Processing and Manufacturing Science 7(2) (1998) 163-172.
[80] K. Kumar, S.V. Kailas, The role of friction stir welding tool on material flow and weld formation,
Materials Science & Engineering A 485(1) (2008) 367-374.
[81] C. Widener, B. Tweedy, D. Burford, Effect of fit-up tolerances on the strenght of friction stir
welds, 47th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference Newport, Rhode Island (2006).
[82] H. Takahara, Y. Motoyama, M. Tsujikawa, S. Oki, S.W. Chung, K. Higashi, Allowance of
Deviation and Gap in Butt Joint on Friction Stir Welding, Advanced Materials Research 15-17 (2007)
375-380.
[83] F. Fadaeifard, K. Matori, S. Abd Aziz, L. Zolkarnain, M. Rahim, Effect of the Welding Speed
on the Macrostructure, Microstructure and Mechanical Properties of AA6061-T6 Friction Stir Butt
Welds, METALS 7(2) (2017) 16.
[84] V. Patel, W. Li, G. Wang, F. Wang, A. Vairis, P. Niu, Friction Stir Welding of Dissimilar
Aluminum Alloy Combinations: State-of-the-Art, METALS 9(3) (2019) 19.
[85] G. Peng, Y. Ma, J. Hu, W. Jiang, Y. Huan, Z. Chen, T. Zhang, Nanoindentation Hardness
Distribution and Strain Field and Fracture Evolution in Dissimilar Friction Stir-Welded AA 6061-
AA 5A06 Aluminum Alloy Joints, ADVANCES IN MATERIALS SCIENCE AND ENGINEERING
2018 (2018).
[86] G. Peng, Q. Yan, J. Hu, P. Chen, Z. Chen, T. Zhang, Effect of Forced Air Cooling on the
Microstructures, Tensile Strength, and Hardness Distribution of Dissimilar Friction Stir Welded
AA5A06-AA6061 Joints (Book review), Metals 9(3) (2019) 304.
[87] Y. Zhou, S.J. Chen, J. Wang, P. Wang, J. Xia, Influences of Pin Shape on a High Rotation Speed
Friction Stir Welding Joint of a 6061-T6 Aluminum Alloy Sheet, METALS 8(12) (2018) 13.
[88] H. Jamshidi Aval, S. Serajzadeh, A study on natural aging behavior and mechanical properties
of friction stir-welded AA6061-T6 plates, The International Journal of Advanced Manufacturing
Technology 71(5) (2014) 933-941.
[89] S. Kang, H. Han, K. Oh, J.-H. Cho, C. Lee, S.-J. Kim, Investigation of the material flow and
texture evolution in friction-stir welded aluminum alloy, Metals and Materials International 15(6)
(2009) 1027-1031.
[90] W. Tao, Z. Yong, L. Xuemei, K. Matsuda, Special grain boundaries in the nugget zone of friction
stir welded AA6061-T6 under various welding parameters, Materials Science & Engineering A 671
(2016) 7-16.
[91] Y. Chen, H. Ding, J.-Z. Li, J.-W. Zhao, M.-J. Fu, X.-H. Li, Effect of welding heat input and post-
welded heat treatment on hardness of stir zone for friction stir-welded 2024-T3 aluminum alloy,
Transactions of Nonferrous Metals Society of China 25(8) (2015) 2524-2532.
[92] A. Polat, M. Avsar, F. Ozturk, Effects of the artificial-aging temperature and time on the
mechanical properties and springback behavior of AA6061, Materials and technology 49(4) (2015)
487-493.
[93] B. Malard, F. De Geuser, A. Deschamps, Microstructure distribution in an AA2050 T34 friction
stir weld and its evolution during post-welding heat treatment, Acta Materialia 101(C) (2015) 90-100.
[94] K. Elangovan, V. Balasubramanian, Influences of post-weld heat treatment on tensile properties
of friction stir-welded AA6061 aluminum alloy joints, Materials Characterization 59(9) (2008) 1168-
1177.
[95] S. Malopheyev, I. Vysotskiy, V. Kulitskiy, S. Mironov, R. Kaibyshev, Optimization of
processing-microstructure-properties relationship in friction-stir welded 6061-T6 aluminum alloy,
Materials Science & Engineering A 662(C) (2016) 136-143.
[96] C. Chang, G. Xiao, E. Liu, J. Lin, X. Zhang, X. Long, L.-L. Zhang, Revisiting the procedure for
characterising mechanical properties in welded joints through nanoindentation, Materials Science and
Technology 35(8) (2019) 986-992.
132
[97] J. Boonma, S. Khammuangsa, K. Uttarasak, J. Dutchaneephet, C. Boonruang, N. Sirikulrat, Post-
Weld Heat Treatment Effects on Hardness and Impact Strength of Aluminum Alloy 6061 Friction
Stir Butt Weld, MATERIALS TRANSACTIONS 56(7) (2015) 1072-1076.
[98] G. Madhusudhan Reddy, P. Mastanaiah, K. Sata Prasad, T. Mohandas, Microstructure and
mechanical property correlations in AA 6061 aluminium alloy friction stir welds, Transactions of the
Indian Institute of Metals 62(1) (2009) 49-58.
[99] K. Krishnan, The effect of post weld heat treatment on the properties of 6061 friction stir welded
joints, Journal of Materials Science 37(3) (2002) 473-480.
[100] H. Aydın, A. Bayram, İ. Durgun, The effect of post-weld heat treatment on the mechanical
properties of 2024-T4 friction stir-welded joints, Materials and Design 31(5) (2010) 2568-2577.
[101] A. Silva, D. Braga, M. Figueiredo, P. Moreira, Ultimate tensile strength optimization of
different FSW aluminium alloy joints, The International Journal of Advanced Manufacturing
Technology 79(5-8) (2015) 805-814.
[102] M. Wang, H. Zhang, J. Zhang, X. Zhang, L. Yang, Effect of Pin Length on Hook Size and Joint
Properties in Friction Stir Lap Welding of 7B04 Aluminum Alloy, Journal of Materials Engineering
and Performance 23(5) (2014) 1881-1886.
[103] Y.N. Zhang, X. Cao, S. Larose, P. Wanjara, Review of tools for friction stir welding and
processing, Canadian Metallurgical Quarterly: The Canadian Journal of Metallurgy and Materials
Science 51(3) (2012) 250-261.
[104] M. Costa, D. Verdera, J.D. Costa, C. Leitao, D.M. Rodrigues, Influence of pin geometry and
process parameters on friction stir lap welding of AA5754-H22 thin sheets, JOURNAL OF
MATERIALS PROCESSING TECHNOLOGY 225 (2015) 385-392.
[105] T. Wang, H. Sidhar, R.S. Mishra, Y. Hovanski, P. Upadhyay, B. Carlson, Effect of hook
characteristics on the fracture behaviour of dissimilar friction stir welded aluminium alloy and mild
steel sheets, Science and Technology of Welding and Joining 24(2) (2019) 178-184.
[106] K. Wang, P. Upadhyay, Y. Wang, J. Li, X. Sun, T. Roosendaal, Investigation of Interfacial
Layer for Friction Stir Scribe Welded Aluminum to Steel Joints, JOURNAL OF
MANUFACTURING SCIENCE AND ENGINEERING-TRANSACTIONS OF THE ASME
140(11) (2018).
[107] V. Gupta, P. Upadhyay, L.S. Fifield, T. Roosendaal, X. Sun, P. Nelaturu, B. Carlson, Linking
process and structure in the friction stir scribe joining of dissimilar materials: A computational
approach with experimental support, Journal of Manufacturing Processes 32(C) (2018) 615-624.
[108] Y. Yue, Z. Li, S. Ji, Y. Huang, Z. Zhou, Effect of Reverse-threaded Pin on Mechanical
Properties of Friction Stir Lap Welded Alclad 2024 Aluminum Alloy, Journal of Materials Science
& Technology 32(7) (2016) 671-675.
[109] M.K. Yadava, R.S. Mishra, Y.L. Chen, B. Carlson, G.J. Grant, Study of friction stir joining of
thin aluminium sheets in lap joint configuration, Science and Technology of Welding and Joining
15(1) (2010) 70-75.
[110] M. Movahedi, A.H. Kokabi, S.M.S. Reihani, H. Najafi, Effect of tool travel and rotation speeds
on weld zone defects and joint strength of aluminium steel lap joints made by friction stir welding,
Science and Technology of Welding and Joining 17(2) (2012) 162-167.
[111] S. Yazdanian, Effect of friction stir lap welding conditions on joint strength of aluminium alloy
6060, IOP Conference Series: Materials Science and Engineering 4(1) (2009) 012021.
[112] S.M. Bayazid, H. Farhangi, A. Ghahramani, Investigation of Friction Stir Welding Parameters
of 6063-7075 Aluminum Alloys by Taguchi Method, Procedia Materials Science 11 (2015) 6-11.
[113] F.C. Liu, Z.Y. Ma, Influence of Tool Dimension and Welding Parameters on Microstructure
and Mechanical Properties of Friction-Stir-Welded 6061-T651 Aluminum Alloy, Metallurgical and
Materials Transactions A 39(10) (2008) 2378-2388.
[114] S. Babu, G.D.J. Ram, P.V. Venkitakrishnan, G.M. Reddy, K.P. Rao, Microstructure and
Mechanical Properties of Friction Stir Lap Welded Aluminum Alloy AA2014, Journal of Materials
Science & Technology 28(5) (2012) 414-426.
133
[115] M. Momeni, M. Guillot, Effect of Tool Design and Process Parameters on Lap Joints Made by
Right Angle Friction Stir Welding (RAFSW), Journal of Manufacturing and Materials Processing
3(3) (2019) 66.
[116] W.R. Longhurst, A.M. Strauss, G.E. Cook, C.D. Cox, C.E. Hendricks, B.T. Gibson, Y.S.
Dawant, Investigation of force-controlled friction stir welding for manufacturing and automation,
Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture
224(6) (2010) 937-949.
[117] G.E. Cook, H. Smartt, J. Mitchell, A.M. Strauss, R. Crawford, Controlling robotic friction stir
welding, Welding Journal 82(28) (2003) 34.
[118] Y.N. Zhang, X. Cao, S. Larose, P. Wanjara, Review of tools for friction stir welding and
processing, Canadian Metallurgical Quarterly 51(3) (2012) 250-261.
[119] M. Balakrishnan, C. Leitão, E. Arruti, E. Aldanondo, D. Rodrigues, Influence of pin
imperfections on the tensile and fatigue behaviour of AA 7075-T6 friction stir lap welds, The
International Journal of Advanced Manufacturing Technology 97(5) (2018) 3129-3139.
[120] E. Maleki, Artificial neural networks application for modeling of friction stir welding effects
on mechanical properties of 7075-T6 aluminum alloy, IOP Conference Series: Materials Science and
Engineering 103 (2015) 012034.
[121] E. Boldsaikhan, E.M. Corwin, A.M. Logar, W.J. Arbegast, The use of neural network and
discrete Fourier transform for real-time evaluation of friction stir welding, Applied Soft Computing
Journal 11(8) (2011) 4839-4846.
[122] M.H. Shojaeefard, R.A. Behnagh, M. Akbari, M.K.B. Givi, F. Farhani, Modelling and Pareto
optimization of mechanical properties of friction stir welded AA7075/AA5083 butt joints using
neural network and particle swarm algorithm, Materials and Design 44 (2013) 190-198.
[123] K.L. Priddy, P.E. Keller, E. Society of Photo-optical Instrumentation, Artificial neural networks
: an introduction, SPIE, Bellingham, Wash., 2005.
134
Appendix: Artificial neural network modeling
Background
The relationship between the tool design, process parameters, and mechanical properties of the FSW
joints is nonlinear and extremely complex. Thus, it is unlikely to have effective analytical expressions
to describe these relations. To optimize the tool design and process parameters, experiments are too
costly and time-consuming. Artificial intelligence (AI) systems such as artificial neural network
(ANN) modeling can be employed to develop the FSW technique while the needed number of
experiments and time is reduced, considerably. The excellent capability of neural networks (NNs)
makes them a great choice to learn a nonlinear mapping from training data sets, recognize the patterns
between input and output parameters without any prior assumptions about their nature, and to
generalize the results. ANN modeling can be employed to model different kinds of manufacturing
techniques such as FSW processes. It is capable of predicting FSW forces and joint mechanical
properties made at different process parameters without considering simplification and assumptions
for the process [73-77, 120-122].
There are numerous researches to predict and optimize the physical and mechanical properties of the
FSW joints using ANN modeling. For example, Okuyucu et al. have studied the relationship between
the tool traverse speed, rotational speed, and mechanical properties of FSW joints for aluminum parts.
A single hidden layer, feed-forward artificial neural network was utilized to study the effect of inputs
on the tensile and yield strength, hardness of HAZ and weld metal, and elongation of the welded
samples [76]. Shojaeefard et al. have evaluated the impact of tool traverse and rotational speed on the
tensile strength and hardness of aluminum joints made by FSW [122]. Buffa et al. have studied the
microhardness and microstructure of FSW joints using ANN modeling [75]. Ghetiya et al. have
investigated the effect of tool shoulder diameter, traverse and rotational speed, and axial force on the
tensile strength of aluminum joints made by FSW. For this purpose, they have developed a
back-propagation neural network [74]. Feed-forward multilayer perceptron was used in most of the
papers on ANN modeling of the FSW process. In these papers, the training algorithm mainly was
back-propagation. In many of these researches, sigmoid transfer function has been used, as it is one
of the best transfer functions for this kind of application.
Theory
A multi-layer perceptron network generally consists of N inputs, M outputs, and L-1 hidden layers
and an output layer, as shown in Figure 93. Adding layers to the perceptron network brings extremely
great modeling power. Indeed, the single-layer perceptron only allows modeling of relationships in
135
which variables are linearly separable. When two layers (a hidden layer and an output layer) are used,
it is possible to model hypersurfaces. For such a two-layer network, the number of neurons in the
hidden layer must be large enough to correctly represent the desired shape. The heuristic below
indicates the way of carrying out learning by back-propagation according to the method of the delta
[123].
Output layer
Hidden layer 2
Hidden layer 1
Input layer
Figure 93: The schematic of multilayer perceptron
The learning algorithm of a multi-layered perceptron (back-propagation) is as follows:
1) Initialize the weights and deviations
w(0) = small random values generally between -1 and 1.
q(0) = small random values generally between -1 and 1.
2) Present the data Xi at the network input and estimate the current output Yest. In the first hidden
layer adjacent to the inputs Xi, we calculate:
I j =  wij X i (D.1)
i
1
Oj = a j = (D.2)
(
1 + exp − ( I j +  j ) 
  )
In the intermediate hidden layers:
I j =  wij Oi (D.3)
i
136
1
Oj = a j = (D.4)
( )
1 + exp − ( I j +  j ) 
 
In the output layer:
I j =  wij Oi (D.5)
i
1
Yest j = a j = (D.6)
( )
1 + exp − ( I j +  j ) 
 
3) Evaluate the necessary correction values in the network from the output error. These delta
values become: In the output layer:
 Sortie j =  Y j (1 − Y j ) (D.7)
 = Ydes j − Yj (D.8)
In the intermediate layers:
 j = O j (1 − O j )  k wkj (D.9)
k
4) Adjust the synaptic coefficients according to the equation:
wij (n + 1) = wij (n) +  j Oi +  ( wij ( n ) − wij ( n − 1) ) (D.10)
Where Oi will take the value Xi on the layer of neurons adjacent to the inputs.
5) Adjust the deviations according to the equation:
 j (n + 1) =  j (n) +  j +  ( j ( n ) −  j ( n − 1) ) (D.11)
6) Repeat steps 2 to 5 for all the data in the training file.
7) Repeat steps 2 to 6 for as many passes as necessary to reduce the sum of the errors in the squared
training data below a prescribed value.
Hidden and output layers with “sigmoid” transfer function were used to predict. The sigmoid transfer
function was:
137
1
F ( x) = (A.12)
1 + e− x
Where x is the weighted sum of the input [123].
Several factors affect the prediction accuracy of the backpropagation neural network models such as
the architecture of the network, momentum coefficient, and learning rate of the model [73, 74, 123].
The factors to compare the different architectures are root-mean-squared error (RMSE), maximum
error, mean relative error (MRE), and mean absolute error (MAE) [76, 123]. Furthermore, the
relationship between the experimental data and predicted values by the developed ANN models has
been studied by calculating the amount of correlation coefficient (R2) using linear regression analysis.
When the correlation coefficient is close to one, it indicates a close relationship between the
experimental data and predicted data by the developed ANN models [73, 76, 77].
Mean relative error:
1 N  Ai − Yi 
MRE =   100 (A.13)
N  Ai 
i
Root-mean squared error:
12
1 N 
RMSE =   ( Ai − Yi )2  (A.14)
N 
 i 
Mean absolute error:
1 N
MAE =  Ai − Yi (A.15)
N
i
Maximum error:
Max.E = max ( Ai − Yi ) (A.16)
Absolute fraction of variance:
138
N 
  ( Ai − Yi )2 
 
R2 = 1 −  i  (A.17)
 N 
  Yi2 
 
 i 
Where Ai, Yi, N, and i are experimental value, predicted value, total number of experimental data, and
trial number, respectively. The formulas are from Ref.[76]
A case study: details of developing ANN models

In this part, the details of developing the ANN models used in chapter 3 is presented. The
experimental data obtained in the previous part of this paper were utilized to train artificial neural
networks in this section. 36 experiments are separated from 40 experiments as training data for the
ANN models; and the 4 remaining tests were utilized as validation tests to evaluate the predictability
and the accuracy of the developed ANN models. Feed-forward neural networks with backpropagation
algorithm were used to model the relationship between the tool geometry and the process parameters
on the downward axial force and the fracture force of the joints. The effect of the architecture of the
network, momentum coefficient, and learning rate of the model on the accuracy of the neural network
models is studied to find the best ANN modeling factors. An ANN model with too small architecture
can result in insufficient degree of freedom, and too large network causes to overfit the data. Thus,
there is an optimized number of neurons and hidden layers to have a reliable ANN model [75]. In this
research, several numbers of neurons in one hidden layer, 4 to 10 neurons, were used to find the
optimal architecture with minimal prediction errors. Table A.1 shows the errors for the validation
data for some single hidden layer ANNs with different numbers of neurons in the hidden layer.
Accordingly, it was found that a single hidden layer ANN with 8 neurons yields the lowest amount
of error without overfitting. Since it shows very good predictability with errors less than 5%, it was
selected for the prediction of fracture forces in this part of the research. The same strategy for
modeling downward axial forces leads to the selection of 10-8-1 architecture as it provides acceptable
predictivity within a small range of errors. Figure 94 illustrates the optimal architecture of the ANN
model for both fracture force and downward axial force. Table 42 indicates the details regarding these
models. The learning rate and momentum both was 0.5. The procedure to select them is indicated in
pages number 95 and 96 in chapter 5. Table 43 shows the input and output for trained and validation
data for fracture force using a 10-8-1 ANN model. Table 44 indicates calculated errors and correlation
coefficients for trained, validation, and the entire set of data for fracture forces modeled by the 10-8-
1 ANN model.
139
Table 41: The calculated errors for the validation data predicted by some single hidden layer ANNs with
different number of neurons in the hidden layer.
Figure 94: The architecture of ANN models in this paper. The architecture of the model for both downward
axial force and fracture force is 10-8-1.
140
Table 42: The details of the developed ANN models.
Welding Force model Failure Force model

Network configuration 10-8-1 10-8-1
Number of inputs 10 10
Number of hidden layers 1 1
Number of neurons 8 8
Number of outputs 1 1
Total No. of experimental data 40 40
Learning rate 0.5
Momentum 0.5
Table 43: the input and output for trained and validation data for fracture force using a 10-8-1 ANN.
Training data Ai Yi
1.8 8.5 0.1 3.5 20 0.25 1700 4250 1.8 1 3118 3116.47
1.8 8.5 0.1 4 20 0.25 1400 3500 1.8 1 2994 2957.13
1.8 8.5 0.1 4 20 0.25 1400 3500 1.85 1 4453 4466.05
1.8 9.5 0.1 4 20 0.45 2000 5000 1.8 2 3390 3487.96
1.8 9.5 0.1 4 20 0.45 2000 5000 1.85 2 3937 3867.06
1.8 10.5 0.25 4.8 24 0.25 1400 3500 1.8 2 5373 5400.5
1.8 10.5 0.25 4.8 24 0.25 1400 3500 1.88 2 5422 5418.97
1.8 11.5 0.25 4.8 24 0.45 2000 5000 1.8 1 2847 2896.17
1.8 11.5 0.25 4.8 24 0.45 2000 5000 1.88 1 5218 5222.88
2.2 8.5 0.1 4.8 24 0.25 2000 5000 2.2 2 3510 3581.13
2.2 8.5 0.1 4.8 24 0.25 2000 5000 2.25 2 3390 3348.62
2.2 9.5 0.1 4.8 24 0.45 1400 3500 2.2 1 5218 5160.85
2.2 9.5 0.1 4.8 24 0.45 1400 3500 2.25 1 5609 5703.94
2.2 10.5 0.25 4 20 0.25 2000 5000 2.2 1 4502 4499.81
2.2 10.5 0.25 4 20 0.25 2000 5000 2.28 1 5756 5774.31
2.2 11.5 0.25 4 20 0.45 1400 3500 2.28 2 3292 3300.49
2.6 8.5 0.25 4 24 0.45 1400 5000 2.6 2 4083 3964.56
2.6 8.5 0.25 4 24 0.45 1400 5000 2.68 2 3314 3425.19
2.6 9.5 0.25 4 24 0.25 2000 3500 2.6 1 3358 3366.38
2.6 10.5 0.1 4.8 20 0.45 1400 5000 2.6 1 6107 5895.4
2.6 10.5 0.1 4.8 20 0.45 1400 5000 2.65 1 5640 5817.04
2.6 11.5 0.1 4.8 20 0.25 2000 3500 2.6 2 2171 2233.35
2.6 11.5 0.1 4.8 20 0.25 2000 3500 2.65 2 2300 2222.77
3 8.5 0.25 4.8 20 0.45 2000 3500 3 1 4982 5005.3
3 8.5 0.25 4.8 20 0.45 2000 3500 3.08 1 6557 6548.64
3 9.5 0.25 4.8 20 0.25 1400 5000 3.08 2 2042 2020.4
3 10.5 0.1 4 24 0.45 2000 3500 3 2 2136 2076.18
3 10.5 0.1 4 24 0.45 2000 3500 3.05 2 2056 2046.48
3 11.5 0.1 4 24 0.25 1400 5000 3 1 3581 3661.85
3 11.5 0.1 4 24 0.25 1400 5000 3.05 1 4141 4107.5
1.75 12 0.3 5 25 0.5 1800 4500 1.75 2 5454 5493.81
1.75 12 0.3 5 25 0.5 1800 4500 1.83 2 5961 5935.44
1.75 12 0.3 4.5 25 0.35 2200 5000 1.75 1 2224 2317.23
1.75 12 0.3 4.5 25 0.35 2200 5000 1.83 1 4043 4084.72
1.75 11.5 0.3 4.5 25 0.45 2400 5500 1.75 2 4328 4275.32
1.75 11.5 0.3 4.5 25 0.45 2400 5500 1.83 2 4582 4652.77
Validation data Ai Yi
1.8 8.5 0.1 3.5 20 0.25 1700 4250 1.835 1 3982 4156.05
2.2 11.5 0.25 4 20 0.45 1400 3500 2.2 2 2709 2809.99
2.6 9.5 0.25 4 24 0.25 2000 3500 2.68 1 4746 4876.42
3 9.5 0.25 4.8 20 0.25 1400 5000 3 2 2123 2024.95
141
Table 44: calculated errors and correlation coefficients for trained, validation, and entire set of data for
fracture forces modeled by 10-8-1 ANN model.
Experimental data Predicted Value Error %
3982 4156 -4.369663486
2709 2810 -3.728313031
4746 4876 -2.739148757
2123 2025 4.616109279
for trained data SSE 183982.4
RMSE 71.48861603
MAX E 211
R2 0.999722461
for validation data SSE 67114.1
RMSE 129.5319459
MAX E 174
R2 0.998734858
for all data SSE 262192.1
RMSE 80.96173479
MAX E 211
R2 0.999634318
MAE 61.845768
MRE 1.73449701
142

Development of A Right Angle Friction Stir Welding (RAFSW) Technique To Assemble Aluminum Products

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Development of A Right Angle Friction Stir Welding (RAFSW) Technique To Assemble Aluminum Products

Transféré par

Droits d'auteur :

Formats disponibles

Development of a right angle friction stir welding

(RAFSW) technique to assemble aluminum products

Doctorat en génie mécanique

© Mahboubeh Momeni, 2020

Sous la direction de:

Michel Guillot, directeur de recherche

Table of Contents ................................................................................................................................ v

List of Figures ..................................................................................................................................... x

List of Tables................................................................................................................................... xvii

Dedication ........................................................................................................................................ xix

Preface .............................................................................................................................................. xxi

General background ........................................................................................................................ 1

Benefits and challenges ................................................................................................................... 1

1.1 FSW Process and the joint specifications.................................................................................. 3

1.1.1 Joint Configuration............................................................................................................. 4

1.1.2 Heat input ........................................................................................................................... 4

1.1.3 Microstructural Evolution .................................................................................................. 5

1.1.4 FSW Joints Defects ............................................................................................................ 7

1.1.5 Residual Stresses ................................................................................................................ 8

1.1.6 Hardness ............................................................................................................................. 9

1.1.7 Tensile Properties ............................................................................................................... 9

1.1.8 Fatigue behavior ............................................................................................................... 10

1.2 Welding parameters................................................................................................................. 11

1.3 Tool Design ............................................................................................................................. 11

1.4 FSW machines ........................................................................................................................ 13

1.5 Making the FSW process affordable ....................................................................................... 16

1.6.2 Effect of Tool Design on Downward Axial Force ........................................................... 18

1.6.3 Effect of Preheating on Axial Force ................................................................................. 19

1.11 Applications of FSW ............................................................................................................. 37

Problems, Objectives, and Methodology ....................................................................... 42

2.1 Problem statement ................................................................................................................... 42

2.1.2 RAFSW ............................................................................................................................ 42

2.1.3 Tool design and process parameters ................................................................................. 43

2.1.4 PWHT .............................................................................................................................. 44

2.1.5 Applications ..................................................................................................................... 44

2.2 Hypothesis ............................................................................................................................... 46

2.3 Objectives of the thesis ........................................................................................................... 46

2.4 General Methodology.............................................................................................................. 47

Development of Friction Stir Welding Technique at Right Angle (RAFSW) Applied on

3.1 Résumé .................................................................................................................................... 50

3.2 Abstract ................................................................................................................................... 50

3.3 Introduction ............................................................................................................................. 51

3.4 Experimental procedure .......................................................................................................... 52

3.4.1 Material properties ........................................................................................................... 53

3.4.3 Metallography .................................................................................................................. 53

3.4.4 Micro-hardness ................................................................................................................. 54

3.4.5 Post weld heat treatment .................................................................................................. 54

3.5 Design of experiments and modeling ...................................................................................... 54

3.5.1 Design of experiments ...................................................................................................... 54

3.5.2 ANN modeling ................................................................................................................. 57

3.6 Results and discussions ........................................................................................................... 60

3.6.1 Effect of tool plunge depth ............................................................................................... 60

3.6.2 Axial force during RAFSW .............................................................................................. 61

3.6.3 Ultimate tensile strength of RAFSW specimens .............................................................. 62

3.6.4 Robustness of the RAFSW process .................................................................................. 63

3.6.5 Effect of artificial aging ................................................................................................... 65

3.6.6 Macrostructure examinations ........................................................................................... 66

3.6.7 Microstructure examinations ............................................................................................ 66

3.6.8 Micro-hardness examinations........................................................................................... 67

Post Weld Heat Treatment effects on Mechanical Properties and Microstructure of

4.1 Résumé .................................................................................................................................... 71

4.2 Abstract ................................................................................................................................... 71

4.3 Introduction ............................................................................................................................. 72