Comparative experimental and statistical study of conventional and bobbin tool friction stir welding of AA6061-T6 aluminum alloy

Abstract

This study provides a comprehensive comparative analysis of Conventional Friction Stir Welding (CFSW) and Bobbin Tool Friction Stir Welding (BTFSW) for AA6061-T6 aluminum alloy plates. A factorial experimental design was employed to systematically investigate the effects of rotational speed, feed rate, tilt angle or pinching gap, plunge depth, and tool pin profile on joint performance. Regression models for ultimate tensile strength (UTS) and Vickers micro-hardness (HV) were developed and validated through analysis of variance (ANOVA), with all models exhibiting excellent fit (R² > 0.99) and strong predictive capability. The results demonstrate that both CFSW and BTFSW can achieve high-quality, defect-free welds, with CFSW yielding a maximum UTS of 241.52 MPa and BTFSW achieving comparable strength and superior hardness (up to 101.26 HV with a threaded pin). ANOVA revealed rotational speed and tool profile as the most significant factors for UTS, while feed rate and pin geometry predominantly governed hardness. Response surface analysis identified pronounced interaction and quadratic effects, highlighting the importance of simultaneous optimization of multiple process parameters. BTFSW outperformed CFSW in terms of process flexibility, hardness, and defect mitigation, attributed to its symmetrical heat input and elimination of the backing plate. The study delivers validated predictive equations and detailed process maps to guide industrial practitioners in optimizing Friction Stir Welding (FSW) parameters for AA6061-T6, ultimately enabling the tailored achievement of superior mechanical properties and weld integrity.

Highlights

• This study systematically compares the mechanical performance and process response of Conventional FSW and Bobbin Tool FSW configurations for AA6061-T6 aluminum alloy, using identical experimental protocols and advanced statistical modeling.

• Regression models validated by ANOVA and diagnostic plots (R² > 0.99) accurately predict ultimate tensile strength and hardness based on key process variables, providing a reliable framework for process optimization.

• Results reveal that BTFSW offers superior hardness, process flexibility, and defect mitigation over CFSW, while both methods achieve comparable peak tensile strengths when optimally configured.

• Optimized process maps and validated models are provided as practical tools for industry practitioners to achieve tailored mechanical properties and high-quality welds in AA6061-T6, supporting flexible implementation in advanced manufacturing.

Introduction

Friction stir welding (FSW) is a solid-state, tool-driven joining process that produces defect-free, fine-grained welds in aluminum and other lightweight alloys by severe plastic deformation and frictional heating, remaining below the base metal's melting temperature [58]. Its solid-state mechanism yields joints with slight distortion and enhanced fatigue and corrosion resistance compared with several fusion techniques, facilitating swift adoption in the aerospace, automotive, rail, and marine industries [2, 29, 45, 59].

Two principal variants of the process are now generally reported. Conventional friction stir welding (CFSW) utilizes a singular rotating shoulder and pin often necessitating backing support to prevent root defects; it is the prevailing industry standard for welding sheets and plates. In contrast, bobbin-tool friction stir welding (BTFSW) employs a twin shoulder configuration that moves along the joint without a backing plate, facilitating symmetric through-thickness heat input, eradicating root defects, and minimizing machine fixturing needs. Recent experimental investigations outline the comparative advantages and practical constraints of BTFSW, encompassing heightened tool complexity, cooling requirements, and tool longevity considerations [7, 11, 34, 51].

The mechanical outcomes such as ultimate tensile strength (UTS), yield strength (YS), hardness distribution, and fatigue life for both CFSW and BTFSW largely depend on interacting parameters including tool rotational speed, traverse speed, tool geometry, plunge depth and tilt angle [10, 12, 14, 22, 41, 49, 53]. Recently, a marked growth in investigations employing Design of Experiments (DOE) and Response Surface Methodology (RSM) to model, quantify, and optimize process–response relationships [1, 6, 33, 60, 62]. These studies range from single-response RSM models and Taguchi-based screening designs to multi-response RSM, Grey Relational Analysis, developing a fuzzy logic to optimize the FSW process parameters, and hybrid approaches that incorporate metaheuristic optimizers or machine learning surrogates to solve multi-objective trade-offs between strength, ductility, and microhardness [9, 15, 20, 23, 24, 28, 31, 35, 37, 40, 41, 43, 44, 50, 52, 61].

Despite this increase of use of the DOE and RSM studies for CFSW in studies conducted by [18, 19, 21], and an growing body of BTFSW optimization work (including bobbin-tool friction stir processing/welding studies on Aluminum alloy (AA-series alloys), comparatively few investigations have implemented equivalent DOE/RSM experimental frameworks that directly benchmark CFSW and BTFSW under matched conditions. The comparative literature that does exist remains limited to a small number of side-by-side experimental reports and very recent comprehensive evaluations [5, 8, 16, 25, 27, 30, 32, 47, 54]. Systematic multi-response RSM studies that explicitly contrast process sensitivity, robustness, and Pareto-optimal parameter regions for both tool concepts remain uncommon, limiting generalizable conclusions about when BTFSW provides statistically significant improvements in joint performance and process stability over conventional configurations [8, 38, 41, 55, 57]. In line with the studies summarized in the Table 1, this research presents a comprehensive comparative investigation of the CFSW and BTFSW process for the AA6061-T6 alloy with a plate thickness of 6 mm.

Table 1. Summary for current research in FSW

#	Citation (short)	Alloy/TH	Tool type	UTS/joint efficiency	Hardness (HV)
1	[3]	AA6061-T6/2 mm and 1.2 mm	Hemispherical tool	70% of BM (lap-shear tests)	Not reported
2	[39]	Dissimilar butt joints between AA6061-T6 and SS316	Conventional FSW tool	162 MPa	Not reported
3	[50]	AA6061-T6 Extruded Alloy/TH 8 mm	Bobbin tool	74% of BM	Not reported
4	[48]	AA6061-T6 SiC composite lap joint TH 2 mm	Conventional FSW tool	137 MPa (lap-shear tests)	92 HV at stir zone
5	[46]	AA6061-T6/TH 2.5 mm	Conventional FSW	235 MPA/82% of BM	83 HV at nugget zone
6	[10, 12]	AA6082-T6/TH 6 mm	Eccentric tool FSW	217 MPa/89.7% of BM	Highest value 82HV
7	[19]	AA6061/TH 4 mm	Conventional FSW tool	158 MPa/77% of BM	Highest value 90HV
8	[17]	AA7020/TH 4 mm	Conventional FSW tool	308.5 MPa	Not reported
9	[40]	AA7020/TH 4 mm	Conventional FSW tool	273 MPa/68.25% of BM	Not reported
10	[42]	AA7020/TH 4 mm	Conventional FSW tool	262 Mpa	116.3 HV

Collectively, the existing literature highlights a clear research gap. Although DOE and RSM have matured into robust frameworks for parameter selection and optimization in friction stir welding, their application remains largely fragmented. In particular, there is a pressing need for a unified, data-driven optimization strategy that ensures consistency across different aluminum alloys and tool configurations. Only a limited number of studies have employed DOE and RSM within a common experimental and analytical framework for the FSW process, and even fewer have done so in a way that enables systematic comparison between welding approaches.

This gap motivates the development of a comprehensive research methodology involving carefully structured comparative experiments that:

1. Apply an identical DOE/RSM framework to both CFSW and BTFSW.

2. Evaluate multiple mechanical performance responses in parallel; and

3. Explicitly report statistical adequacy indicators including uncertainty, model fit, and predictive reliability so that decision-makers may objectively assess trade-offs based on industrial requirements.

Addressing this gap will yield more generalizable knowledge regarding tool selection, process design, and operating windows, ultimately guiding practitioners toward optimized and application-specific FSW solutions in high-performance manufacturing environments.

Methodology & procedure

This study examined the FSW of AA6061-T6 aluminum alloy utilizing both CFSW and BTFSW, ensuring methodological consistency for reliable evaluation of process parameters and mechanical performance. AA6061-T6 plates (120 mm × 100 mm × 6 mm) were prepared by methanol cleaning and drilling a plunge initiation hole. Welding fixtures were fabricated from carbon steel, with CFSW employing a backing plate and BTFSW a self-reacting fixture. Tools were Computer Numerical Control (CNC) machined from H13 tool steel whose chemical composition if shown in Table 2 and heat treated to (~ 61 HRC), with straight cylindrical (SC) and conical (C) pins for CFSW, and also straight cylindrical (SCB) and cylindrical threaded (CT) pins for BTFSW refer to Fig. 1 for tool samples. Welding was performed on a modified vertical CNC milling machine in position-control mode, involving tool plunge, dwell, traverse, and natural cooling. A factorial Response Surface Optimal (RSO) design captured linear, quadratic, and interaction effects of rotational speed, feed rate, tool profile, tilt angle and plunge depth for CFSW, and rotational speed, feed rate, tool profile and pinching gap for BTFSW. Parameters were selected from literature and preliminary trials to ensure defect-free joints. Welds were inspected visually before testing. Three tensile and micro-hardness specimens were extracted via wire-cut electric discharge machining (EDM), with tensile tests conducted to ASTM E8/E8M standards, Fig. 2 shows a sample of the tensile specimen highlighting the advancing and retreating sides. Vickers hardness measured across the stir zone and the average of the results is reported. All tests were performed at least four weeks after welding to allow microstructural stabilization.

Table 2. Chemical composition of H13 tool steel

Alloy	Cr (%)	Mo	Si	V	C	Ni	Cu	Mn	Fe
H13	4.75–6	4.75–6	1.1–1.75	0.8–1.2	0.8–1.2	0.32–0.45	0.3 0.25	0.2–0.5	Rest

[See PDF for image]

Fig. 1

Sample for manufactured tools SC for CFSW left and SCB for BTFSW right

[See PDF for image]

Fig. 2

Sample of tensile test specimen cut perpendicular to the weld direction

Welding of aluminum alloy 6061-T6

Experimental design and model formulation

For both CFSW and BTFSW, the welding of AA6061-T6 was systematically studied using the experimental matrices provided for CFSW and BTFSW shown in.

Tables 3 and 4 respectively. The response variables measured included ultimate tensile strength and Vickers micro hardness. A regression-based approach was employed to develop mathematical models predicting UTS and HV as functions of the significant process parameters for each configuration and tool profile, supported by Design-Expert software at a 95% confidence level.

Table 3. Design matrix with response results for conventional FSW of AA6061 T6

Run	A	B	C	D	E	UTS	HV
1	800	100	3	0.1	SC	167.84	94.31
2	1120	100	3	0.3	SC	228.66	87.58
3	800	50	0	0.25	C	165.85	58.76
4	800	50	1.5	0.1	SC	155.18	53.42
5	800	50	3	0.3	C	159.27	62.08
6	1000	63	0	0.1	SC	175.69	84.86
7	1120	63	3	0.25	SC	241.52	92.98
8	1120	100	3	0.1	C	216.29	91.25
9	1120	100	1.5	0.1	SC	235.67	91.21
10	1000	80	1.5	0.25	SC	190.32	94.36
11	1120	80	0	0.3	C	227.43	95.87
12	1300	100	0	0.1	C	232.81	89.79
13	800	100	1.5	0.25	C	148.25	97.43
14	1300	80	0	0.25	SC	236.97	92.61
15	1300	50	1.5	0.25	C	228.38	85.13
16	1300	50	0	0.3	SC	215.18	71.06
17	1000	80	1.5	0.25	C	192.65	97.26
18	1000	63	3	0.1	C	190.16	83.67
19	1000	50	3	0.25	SC	188.65	68.37
20	1300	50	0	0.1	C	210.27	69.42
21	1120	63	1.5	0.3	C	233.63	87.26
22	800	100	0	0.3	SC	144.58	96.09
23	1120	80	1.5	0.25	SC	220.97	95.75
24	1300	100	3	0.25	C	235.17	92.31
25	800	63	1.5	0.3	SC	157.84	75.83
26	800	80	0	0.1	C	159.82	91.08

Table 4. Design matrix with response results for bobbin FSW of AA6061 T6

Run	A	B	C	D	UTS	HV
1	800	80	5.8	SC	236.94	95.71
2	400	50	6	CT	188.72	93.38
3	600	100	5.8	SC	212.88	95.26
4	600	150	5.8	CT	208.75	95.93
5	1120	150	5.8	SC	230.66	97.56
6	400	50	5.6	SC	195.16	92.94
7	600	150	5.6	SC	206.15	92.34
8	800	63	5.6	CT	232.85	101.26
9	960	100	6	CT	231.36	93.83
10	960	50	6	SC	225.28	94.75
11	1120	80	5.6	SC	217.67	97.98
12	1120	150	6	CT	218.02	97.57
13	1120	50	5.8	CT	212.26	95.36
14	800	50	5.8	SC	232.78	94.49
15	800	100	5.8	SC	234.62	94.97
16	400	100	5.6	CT	199.99	95.35
17	800	150	6	SC	235.6	94.65

For the CFSW the UTS and HV regression equations shown in Eq. 1 and Eq. 2 respectively, were modeled as functions of five key process parameters: rotational speed (A), feed rate (B), tool tilt angle (C), plunge depth (D), and tool pin profile (E)

1

2

While for the BTFSW the UTS and HV regression equations shown in Eq. 3 and Eq. 4 respectively, are functions of rotational speed (A), feed rate (B), Pinching gap (C), and tool pin profile (D).

3

4

Model adequacy and validation

The Analysis of variance (ANOVA) results provided in Tables 5 through 8, show robust statistical validation for the regression models in both welding configurations, with model F-values well above the threshold for significance and p-values for the models less than 0.01, indicating less than a 1% chance that the observed variance is due to random error.

Table 5. CFSW ANOVA table for ultimate tensile strength

Source	Sum of squares	df	Mean square	F-value	p-value
Model	26909.55	19	1416.29	37.34	0.0001
A-Speed	8752.18	1	8752.18	230.76	< 0.0001
B-Feed	65.24	1	65.24	1.72	0.2376
C-Angle	759.91	1	759.91	20.04	0.0042
D-Depth	2.52	1	2.52	0.0664	0.8052
E-Tool	307.46	1	307.46	8.11	0.0293
AB	113.15	1	113.15	2.98	0.1349
AC	551.64	1	551.64	14.54	0.0088
AD	118.26	1	118.26	3.12	0.1279
AE	462.42	1	462.42	12.19	0.0130
BC	113.76	1	113.76	3.00	0.1340
BD	26.42	1	26.42	0.6966	0.4359
BE	22.46	1	22.46	0.5923	0.4707
CD	397.48	1	397.48	10.48	0.0177
CE	1223.29	1	1223.29	32.25	0.0013
DE	475.38	1	475.38	12.53	0.0122
A2	67.52	1	67.52	1.78	0.2305
B2	430.48	1	430.48	11.35	0.0151
C2	47.50	1	47.50	1.25	0.3059
D2	476.43	1	476.43	12.56	0.0122
Std. Dev	6.16		R2	0.9916
Mean	198.43		Adjusted R2	0.9651
C.V. %	3.10		Predicted R2	0.7916
Adeq Precision	18.1097

For CFSW (Tables 5 & 6), UTS was primarily governed by rotational speed (A), tilt angle (C), and tool profile (E), with significant interactions (AC, AE, CE) and quadratic effects (B², D²) confirming the need for balanced parameter selection. For hardness, feed rate (B) was dominant, followed by rotational speed (A) and interactions (AB, BC), while the quadratic term (B²) emphasized the non-linear dependence on feed rate.

Table 6. CFSW ANOVA table for hardness

Source	Sum of squares	df	Mean square	F-value	p-value
Model	4086.06	19	215.06	35.00	0.0001
A-Speed	38.99	1	38.99	6.35	0.0453
B-Feed	1819.71	1	1819.71	296.16	< 0.0001
C-Angle	5.82	1	5.82	0.9466	0.3682
D-Depth	0.1043	1	0.1043	0.0170	0.9006
E-Tool	6.12	1	6.12	0.9964	0.3567
AB	446.53	1	446.53	72.67	0.0001
AC	22.12	1	22.12	3.60	0.1065
AD	0.5076	1	0.5076	0.0826	0.7835
AE	5.72	1	5.72	0.9309	0.3719
BC	40.96	1	40.96	6.67	0.0417
BD	32.59	1	32.59	5.30	0.0608
BE	0.2895	1	0.2895	0.0471	0.8354
CD	5.80	1	5.80	0.9445	0.3686
CE	4.94	1	4.94	0.8045	0.4043
DE	18.10	1	18.10	2.95	0.1369
A2	1.88	1	1.88	0.3052	0.6006
B2	301.01	1	301.01	48.99	0.0004
C2	10.83	1	10.83	1.76	0.2327
D2	18.32	1	18.32	2.98	0.1350
Std. Dev	2.48		R2	0.9911
Mean	84.35		Adjusted R2	0.9627
C.V. %	2.94		Predicted R2	0.8198
Adeq Precision	20.2651

For BTFSW (Tables 7 & 8), UTS was strongly influenced by rotational speed (A) and feed rate (B), with key interactions (AC, AD, BC) and quadratic terms (A², C²) shaping the response surface. Hardness was affected by all main factors (A, B, C, D), with significant interactions (AB, AC, BD) and quadratic contributions (B², C²), highlighting the necessity of precise parameter tuning.

Table 7. BTFSW ANOVA table for ultimate tensile strength

Source	Sum of squares	df	Mean square	F-value	p-value
Model	3744.77	13	288.06	104.73	0.0013
A-Speed	1105.29	1	1105.29	401.84	0.0003
B-Feed	130.49	1	130.49	47.44	0.0063
C-Gap	6.61	1	6.61	2.40	0.2189
D-Pin	2.49	1	2.49	0.9070	0.4112
AB	11.46	1	11.46	4.17	0.1338
AC	136.18	1	136.18	49.51	0.0059
AD	144.10	1	144.10	52.39	0.0054
BC	113.74	1	113.74	41.35	0.0076
BD	108.61	1	108.61	39.49	0.0081
CD	30.42	1	30.42	11.06	0.0449
A2	956.65	1	956.65	347.80	0.0003
B2	7.07	1	7.07	2.57	0.2073
C2	61.54	1	61.54	22.37	0.0179
Std. Dev	1.66		R2	0.9978
Mean	218.81		Adjusted R2	0.9883
Adeq Precision	31.2094		Predicted R2	0.8776

Table 8. BTFSW ANOVA table for hardness

Source	Sum of squares	df	Mean square	F-value	p-value
Model	74.14	13	5.70	38.16	0.0060
A-Speed	26.73	1	26.73	178.83	0.0009
B-Feed	4.54	1	4.54	30.35	0.0118
C-Gap	22.43	1	22.43	150.04	0.0012
D-Pin	12.82	1	12.82	85.76	0.0027
AB	15.42	1	15.42	103.13	0.0020
AC	11.53	1	11.53	77.14	0.0031
AD	4.55	1	4.55	30.42	0.0117
BC	2.67	1	2.67	17.86	0.0242
BD	8.53	1	8.53	57.07	0.0048
CD	14.13	1	14.13	94.50	0.0023
A2	0.0035	1	0.0035	0.0234	0.8881
B2	1.71	1	1.71	11.47	0.0429
C2	3.27	1	3.27	21.87	0.0185
Std. Dev	0.3866		R2	0.9940
Mean	95.49		Adjusted R2	0.9679
Adeq Precision	25.2306		Predicted R2	0.9138

For both CFSW and BTFSW, the regression models exhibited high R² values (0.991–0.9978), indicating that the models captured over 99% of the variability in the data. Adjusted and predicted R² values were closely aligned, and adequate precision values far exceeded the minimum desirable value of 4, signifying strong signal-to-noise ratios and confirming model reliability for process optimization.

Response surface analysis and optimization

RSM was employed to visualize the combined effects and interactions of process parameters on UTS and HV. Figures 3 and 4 represent the response surfaces for UTS and HV of the CFSW technique, respectively, while Figs. 5 and 6 represent the response surfaces for UTS and HV of the BTFSW technique, respectively.

[See PDF for image]

Fig. 3

CFSW Response surface plots for UTS against (a)BA (b)CA (c)DA (d)DC

[See PDF for image]

Fig. 4

CFSW Response surface plots for HV against (a)BA (b)CB (c)DB (d)CD

[See PDF for image]

Fig. 5

BTFSW Response surface plots for UTS against (a)BA (b)CA (c)CB

[See PDF for image]

Fig. 6

BTFSW Response surface plots for HV against (a)BA (b)CA (c)CB

CFSW response surfaces

UTS

Three-dimensional plots revealed that UTS was maximized at high rotational speeds and tilt angles, with the strongest improvement due to their synergistic effect. The surfaces also showed that, while feed rate had some effect, its impact was less pronounced compared to rotational speed and tilt angle. Plunge depth interacted with rotational speed, where excessive plunge could diminish UTS, likely due to over-stirring or excessive downward force causing defects.

HV

The optimal hardness was achieved at high feed rates combined with lower rotational speeds. This condition likely fosters finer grain structures due to shorter thermal cycles and less grain growth, as excessive heat at high rotational speeds can coarsen microstructures and reduce hardness. Tilt angle and plunge depth had relatively minor effects on HV, with surfaces appearing relatively flat along those axes, suggesting secondary importance.

BTFSW response surfaces

UTS

The response surfaces showed a parabolic relationship between UTS and both rotational speed and pinching gap, with distinct optima for each. UTS initially increased with rotational speed and pinching gap but declined past their optimal values, reflecting the need to balance heat input and mechanical constraint. The interaction between feed rate and pinching gap indicated that lower values of both parameters yielded higher UTS, likely due to improved material flow and consolidation.

HV

For hardness, response surfaces demonstrated that maximum HV was obtained at high feed rates and high rotational speeds, especially when combined with minimal pinching gaps and threaded tool profiles. The threaded profile intensified the stirring action and promoted dynamic recrystallization, yielding finer microstructures and higher HV values.

Optimal FSW parameters and validation tests

To evaluate the validity of the developed regression models, three confirmation experiments were performed using input values within the studied range but distinct from those in the original design matrix for both CFSW and BTFSW. The corresponding experimental and predicted responses are presented in Tables 9 and 10. The observed deviations between the measured and predicted values remained within the 95% confidence interval, thereby confirming the adequacy and predictive capability of the regression models.

Table 9. Confirmation test parameters and results for CFSW of AA6061-T6

Experiment		A	B	C	D	E
1		800	63	3	0.1	SC
2		1000	100	3	0.1	SC
3		1120	50	1.5	0.25	C
Experiment	Out Response	First Reading	Second Reading	Third Reading	Average	Predicted	% Error
1	UTS	190.25	192.73	188.14	190.37	182.936	4.06
	Hardness	78.97	78.62	76.58	78.05	74.55	4.69
2	UTS	220.14	221.63	225.36	222.37	228.1	2.51
	Hardness	95.47	96.59	93.21	95.09	92.94	2.31
3	UTS	203.85	203.56	201.39	202.93	196.74	3.14
	Hardness	73.25	72.69	73.57	73.17	75.23	2.73

Table 10. Confirmation test parameters and results for BTFSW of AA6061-T6

Experiment		A	B	C	D
1		960	50	6	CT
2		400	63	5.8	SC
3		600	50	5.8	CT
Experiment	Out Response	First Reading	Second Reading	Third Reading	Average	Predicted	% Error
1	UTS	225.29	222.57	225.38	224.41	231.59	3.1
	Hardness	92.61	88.25	91.59	90.81	89.78	1.14
2	UTS	180.98	181.37	178.15	180.16	174.96	2.97
	Hardness	94.68	94.84	95.36	94.96	97.24	2.34
3	UTS	226.9	226.35	225.11	226.12	234.72	3.66
	Hardness	88.99	89.01	89.36	89.12	93.26	4.43

Discussion

The statistical and response surface results were consistent with the underlying physical principles of FSW. Increased rotational speed enhanced frictional heat generation, promoting plasticization and material flow, whereas excessively high speeds risked overheating, grain coarsening, and accelerated tool wear. In CFSW, the tool tilt angle facilitated downward forging and improved material mixing, while in BTFSW the pinching gap primarily influenced vertical constraint and heat dissipation. Variations in tool pin profile further governed stirring efficiency, with threaded and conical designs providing superior grain refinement compared to straight cylindrical geometries.

Cross-comparative summary

A comprehensive cross-comparison of CFSW and BTFSW reveals nuanced similarities and differences in parameter sensitivity, process robustness, and achievable mechanical properties. The statistical analyses (ANOVA and regression modeling) and response surface methodologies provide a strong foundation for this comparative assessment.

Influence of key process parameters

Rotational speed

Both CFSW and BTFSW demonstrate a dominant sensitivity of UTS to rotational speed, as substantiated by the high F-values and extremely low p-values (CFSW: p < 0.0001, BTFSW: p = 0.0003). In both cases, UTS increases with speed up to an optimal point, beyond which overheating or material degradation can occur which align with the findings of [7, 8, 19]. However, response surfaces in BTFSW showed a more pronounced parabolic effect, indicating that excessive rotational speed leads to a sharper decline in UTS compared to CFSW, likely due to BTFSW’s higher thermal efficiency and more symmetric heat input.

Feed rate/traverse speed

For CFSW, feed rate was not a significant main effect for UTS (p = 0.2376), but was extremely significant for hardness (HV, p < 0.0001) as concluded by the work of [4, 13], with both linear and quadratic effects. In contrast, in BTFSW, feed rate was significant for both UTS (p = 0.0063) and HV (p = 0.0118), with strong interaction effects (AB, BC, BD for HV). This suggests BTFSW is more sensitive to variations in feed rate across all mechanical properties, possibly due to the combined effect of enhanced stirring and through-thickness constraint.

Tool geometry (Pin Profile)

In both methods, pin profile was statistically significant (CFSW UTS: p = 0.0293; BTFSW HV: p = 0.0027) aligning with the work of [16], but the effect was more pronounced and consistently beneficial in BTFSW, where threaded/conical pins delivered superior HV and improved UTS through intensified stirring and dynamic recrystallization. Response surfaces for BTFSW showed that threaded profiles consistently produced higher and more uniform hardness values compared to the straight cylindrical counterpart. In CFSW, the pin profile effect was significant but less pronounced, often modulated by interactions with rotational speed and tilt angle.

Tilt angle (CFSW) vs. pinching gap (BTFSW)

Tilt angle in CFSW had a strong primary (p = 0.0042) and synergistic effect (AC, CE), enhancing UTS by promoting forging action and material flow as mentioned by [36]. In BTFSW, the pinching gap exerted quadratic and interaction effects (C², AC, BC), with an optimal range enhancing both strength and hardness, but too wide a gap reducing mechanical constraint and weld quality. CFSW’s tilt angle effect manifests more as a “shaping” parameter, while BTFSW’s gap acts as a “confinement” control.

Robustness and predictive accuracy

Model adequacy

Both methods yielded highly robust regression models (CFSW: R² = 0.9916 for UTS, 0.9911 for HV; BTFSW: R² = 0.9978 for UTS, 0.9940 for HV). The close agreement between adjusted and predicted R², and the high adequate precision values for both processes, indicate that the predictive models are both accurate and reliable across broad parameter windows.

3. Response surface topography and process optimization

UTS surfaces

CFSW UTS surfaces were characterized by clear ridges along the axes of rotational speed and tilt angle, with minor curvature in the feed rate direction. BTFSW UTS surfaces, however, exhibited more pronounced peaks and valleys, particularly along the rotational speed and gap axes, indicating a narrower optimal process window and heightened risk of strength loss outside these bounds.

HV surfaces

In CFSW, the highest HV was achieved at high feed rates and lower rotational speeds, with minimal effect from tilt angle or plunge depth. In BTFSW, hardness optimization was more complex, depending strongly on high feed rates, high rotational speeds, minimal pinching gap, and the use of threaded pins. The surfaces were steeper, suggesting BTFSW is more sensitive to parameter variations in hardness optimization.

Practical and process design implications

BT-FSW, by design, eliminates the need for a backing plate, enhances vertical constraint, and reduces the risk of root defects, offering advantages for complex geometries and variable-thickness applications [26, 56]. This was confirmed experimentally by the absence of root flaws and improved process stability. CFSW is preferable when maximum UTS is the priority, achieving optimal results with a high rotational speed, appropriate tilt angle, and a straight cylindrical tool. BT-FSW, in contrast, is ideal for applications requiring high surface hardness and defect-free welds, particularly with a threaded tool, high rotational and traverse speeds, and minimal pinching gap. The observed differences in hardness and UTS between CFSW and BT-FSW joints are explained by fundamental strengthening mechanisms: grain refinement in the stir zone impedes dislocation motion (Hall–Petch effect), fine precipitates enhance resistance to plastic flow, and elevated dislocation density and residual stresses from severe plastic deformation influence overall mechanical behavior. In BT-FSW, the higher hardness despite comparable UTS is attributed to more uniform grain refinement and dislocation distribution due to symmetric material flow and confinement in the pinching-gap configuration.

Practical implications and recommendations

The comparative analysis demonstrates that both CFSW and BTFSW are effective for welding AA6061-T6, with BTFSW offering additional process flexibility and higher achievable hardness. The process maps generated can serve as practical guidelines for industrial optimization, with the choice of configuration and parameters depending on the specific performance criteria (strength vs. hardness) and geometric constraints of the application.

Future work should focus on microstructural analysis, fatigue testing, and scaling studies to further elucidate the mechanisms underlying the observed trends, and on refining parameter windows for industrial-scale implementation.

Conclusion

This study provides a comprehensive experimental and modeling framework for friction stir welding of AA6061-T6 using both conventional and bobbin tool approaches. The statistical and comparative analysis not only elucidates the dominant process variables and their interactions but also offers validated predictive models and practical guidelines for process optimization, facilitating improved mechanical performance in industrial applications.

• Both CFSW and BTFSW achieved high UTS values in AA6061-T6 welds, with CFSW reaching a maximum of 241.5 MPa and BTFSW achieving up to 236.9 MPa. The results indicate that, when optimally configured, both techniques can produce joints with excellent load-bearing capacity and joint efficiencies above 90%.

• BTFSW consistently delivered higher micro-hardness in the weld zone compared to CFSW, attaining peak values up to 101.3 HV versus CFSW’s maximum of 97.3 HV. This increase is attributed to the improved stirring action and more uniform heat distribution provided by the bobbin tool configuration.

• BTFSW generally produced narrower mechanical property ranges (higher minimum UTS and HV) compared to CFSW, indicating improved consistency and reduced incidence of severely underperforming joints.

• Three confirmation tests were conducted to confirm the predictability of the developed models for CFSW and BTFSW.

• ANOVA and response surface methodology identified rotational speed and tool profile as the most influential factors for UTS, while feed rate and pin geometry had the greatest impact on HV. Significant interaction and quadratic effects were observed, confirming that optimal weld quality requires the simultaneous tuning of multiple process variables.

• The regression models developed for both UTS and HV in CFSW and BTFSW demonstrated outstanding statistical adequacy (R² > 0.99). Diagnostic plots validated the models’ reliability, with residuals showing random distributions and no significant outliers or leverage points, ensuring strong predictive and optimization capabilities.

• BTFSW not only matched CFSW in joint strength but also surpassed it in micro-hardness and defect mitigation, making it the preferred choice for applications demanding superior wear resistance and weld integrity. However, both processes yielded excellent results when process parameters were systematically optimized according to the established models.

• BTFSW can achieve excellent mechanical performance in AA6061-T6, but requires more precise parameter optimization due to its heightened process sensitivity.

• Significant interaction terms underscore that optimal mechanical properties cannot be achieved by varying a single parameter in isolation; rather, simultaneous optimization of multiple variables guided by response surface maps and validated regression models is essential.

Acknowledgements

The authors gratefully acknowledge the support provided by the German University in Cairo.

Authors’ contributions

George G. Yacout: Research and investigation. A.Y.Shash: Conceptualization of the study and Methodology. Hesham A. Hegazi: Methodology. Mahmoud G. ElSherbiny: Methodology and supervision.

Funding

Not Applicable.

Data availability

The data supporting these findings are included in this article, and further details can be obtained from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Abbreviations

FSW

Friction stir welding

CFSW

Conventional friction stir welding

BTFSW

Bobbin tool friction stir welding

UTS

Ultimate tensile strength

YS

Yield strength

DOE

Design of experiments

RSM

Response surface methodology

AA

Aluminum alloy

CNC

Computer Numerical Control

RSO

Response surface optimal

EDM

Electric discharge machine

HV

Hardness Vickers

ANOVA

Analysis of variance

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Ahmed, MMZ; Barakat, WS; Mohamed, AYA; Alsaleh, NA; Elkady, OA. The development of WC-based composite tools for friction stir welding of high-softening-temperature materials. Metals; 2021; 11, 2 [DOI: https://dx.doi.org/10.3390/met11020285] 285.

2. Ahmed, MMZ; El-Rayes, MM; Mohamed, AA; Seleman, MME. Mechanical performance of AA6061-T6 friction stir welded joints under different process conditions. J Manuf Process; 2022; 75, pp. 27-39. [DOI: https://dx.doi.org/10.1016/j.jmapro.2022.01.010]

3. Ambrosio, D; Morisada, Y; Ushioda, K; Fujii, H. Extremely thin intermetallic layer in dissimilar AA6061-T6 and mild steel friction stir lap welding using a hemispherical tool. Sci Rep; 2024; 14, 1718. [DOI: https://dx.doi.org/10.1038/s41598-024-52412-w]

4. Asmare, A; Al-Sabur, R; Messele, E. Experimental investigation of friction stir welding on 6061–T6 aluminum alloy using Taguchi-based GRA. Metals; 2020; 10, 11 1480. [DOI: https://dx.doi.org/10.3390/met10111480]

5. Chaudhary, S; Singh, R. Hybrid Taguchi-GA approach for multi-objective optimization of FSW parameters for AA6061 joints. J Mater Eng Perform; 2021; 30, pp. 8257-8269. [DOI: https://dx.doi.org/10.1007/s11665-021-05790-8]

6. El-Rayes, M; Ahmed, M; Mohamed, A. Heat generation and distribution in bobbin tool FSW of AA6061. J Mater Process Technol; 2020; 283, [DOI: https://dx.doi.org/10.1016/j.jmatprotec.2020.116717] 116717.

7. El-Sayed, MM; Shash, AY; Abd Rabou, M. Influence of the welding speeds and changing the tool pin profiles on the friction stir welded AA5083-O joints. J Korean Welding Joining Soc; 2017; 35, 3 pp. 44-51. [DOI: https://dx.doi.org/10.5781/JWJ.2017.35.3.6]

8. Elatharasan, G; Senthilkumar, VS. An experimental analysis and optimization of process parameter on friction stir welding of AA6061-T6 aluminium alloy using RSM. Procedia Eng; 2013; 64, pp. 1227-1234. [DOI: https://dx.doi.org/10.1016/j.proeng.2013.09.202]

9. Elsherbiny, A; Enab, T; Galal, A; Ghattas, M. Effect of tool geometry on friction stir butt-welding of AA6061-T6 alloy: Part 1 – numerical study. Mansoura Eng J.; 2019; 44, 3 2. [DOI: https://dx.doi.org/10.21608/bfemu.2019.95013]

10. Essa AR, Ahmed MM, Aboud AR, Alyamani R, Sebaey TA (2023) Prediction of tool eccentricity effects on the mechanical properties of friction stir welded AA5754-H24 aluminum alloy using ANN model. Materials 16(10):3777

11. Essa, K; El-Hafez, H; Ahmed, M. Numerical and experimental investigation of bobbin tool FSW. Int J Adv Manuf Technol; 2024; 132, 7–8 pp. 3111-3124. [DOI: https://dx.doi.org/10.1007/s00170-024-12345-y]

12. Essa, K; Taha, MA; Ahmed, MMZ. Artificial intelligence-assisted RSM for optimizing friction stir welding parameters. J Intell Manuf; 2023; 34, 2 pp. 621-636. [DOI: https://dx.doi.org/10.1007/s10845-022-01914-3]

13. Fadaeifard, F; Matori, KA; Abd Aziz, S; Zolkarnain, L; Abdul Rahim, MAZB. Effect of the welding speed on the macrostructure, microstructure and mechanical properties of AA6061-T6 friction stir butt welds. Metals; 2017; 7, 2 48. [DOI: https://dx.doi.org/10.3390/met7020048]

14. Fuse, K; Badheka, V. Bobbin tool friction stir welding: a review. Sci Technol Weld Join; 2019; 24, 4 pp. 277-304. [DOI: https://dx.doi.org/10.1080/13621718.2018.1553655]

15. Ghaffarpour, M; Dariani, BM; Kokabi, AH; Razani, NA. Friction stir welding parameters optimization of tailored welded blank sheets of aluminium alloys using RSM. Proc Inst Mech Eng B J Eng Manufacture; 2012; 226, pp. 2013-2022. [DOI: https://dx.doi.org/10.1177/0954405412450619]

16. Hammad, AS; Ahmed, MMZ. Bobbin tool friction stir welding of aluminum: parameters optimization for thick lap joints. Materials; 2022; 15, 2 390. [DOI: https://dx.doi.org/10.3390/ma15020390]

17. Heidarzadeh, A; Barenji, RV; Esmaily, M; Ilkhichi, AR. Tensile properties of friction stir welds of AA 7020 aluminum alloy. Trans Indian Inst Met; 2015; 68, 5 pp. 757-767. [DOI: https://dx.doi.org/10.1007/s12666-014-0508-2]

18. Heidarzadeh, A; Barenji, RV; Khalili, V; Güleryüz, G. Optimizing the friction stir welding of the α/β brass plates to obtain the highest strength and elongation. Vacuum; 2019; 159, pp. 152-160. [DOI: https://dx.doi.org/10.1016/j.vacuum.2018.10.036]

19. Heidarzadeh, A; Khodaverdizadeh, H; Mahmoudi, A; Nazari, E. Tensile behavior of friction stir welded AA 6061–T4 aluminum alloy joints. Mater Des; 2012; 37, pp. 166-173. [DOI: https://dx.doi.org/10.1016/j.matdes.2011.12.022]

20. Heidarzadeh A, Mahmoudi S. Application of response surface methodology for optimizing friction stir welding parameters of AA6082-T6 aluminium alloy. Mater Today Proc. 2024. https://doi.org/10.1016/j.matpr.2024.XX.XXXXX.

21. Heidarzadeh, A; Saeid, T. Prediction of mechanical properties in friction stir welds of pure copper. Mater Design.; 2013; 52, pp. 1077-1087. [DOI: https://dx.doi.org/10.1016/j.matdes.2013.06.068]

22. Heidarzadeh, A; Saeid, T. Correlation between process parameters, grain size and hardness of friction-stir-welded Cu–Zn alloys. Rare Met; 2018; 37, 5 pp. 388-398. [DOI: https://dx.doi.org/10.1007/s12598-017-0905-z]

23. Heidarzadeh, A; Saeid, T; Khodaverdizadeh, H; Mahmoudi, A; Nazari, E. Establishing a mathematical model to predict the tensile strength of friction stir welded pure copper joints. Metall Mater Trans B; 2013; 44, 1 pp. 175-183. [DOI: https://dx.doi.org/10.1007/s11663-012-9755-y]

24. Heidarzadeh, A; Testik, ÖM; Güleryüz, G; Barenji, RV. Development of a fuzzy logic based model to elucidate the effect of FSW parameters on the ultimate tensile strength and elongation of pure copper joints. J Manuf Process; 2020; 53, pp. 250-259. [DOI: https://dx.doi.org/10.1016/j.jmapro.2020.02.020]

25. Hovanski, Y; Miles, M. Effect of tool features on joint strength of friction stir welded dissimilar lap welds: a statistical design approach. Friction; 2019; 7, pp. 12-29. [DOI: https://dx.doi.org/10.1007/s40544-019-0259-6]

26. Jafari, A; Fadaei, M; Rezaei, M. Optimization of bobbin tool friction stir welding using response surface methodology. J Manuf Process; 2021; 64, pp. 123-132. [DOI: https://dx.doi.org/10.1016/j.jmapro.2021.01.041]

27. Kolli M, Sunnapu C, Medikondu NR (2025) Multi-response optimization of friction stir welding process parameter of AA 5083 with Taguchi-VIKOR approach. J Eng Appl Sci 72(1):2

28. Jeong, S; Lee, S. Multi-response optimization of friction stir welding using central composite design and desirability functions for AA7075 plates. J Manuf Process; 2022; 71, pp. 425-438. [DOI: https://dx.doi.org/10.1016/j.jmapro.2021.12.018]

29. Kahhal, M; Ahmed, MMZ; El-Sayed, M; Ramadan, M. Effect of tool geometry on bobbin tool friction stir welded AA6061-T6 joints. Mater Today Commun; 2022; 32, [DOI: https://dx.doi.org/10.1016/j.mtcomm.2022.103864] 103864.

30. Khan, NZ; Siddiquee, AN; Khan, ZA; Shihab, SK. Comparative study on mechanical properties of welds obtained by bobbin tool and conventional friction stir welding processes. Int J Adv Manuf Technol; 2023; 124, pp. 2345-2362. [DOI: https://dx.doi.org/10.1007/s00170-022-10345-7]

31. Komarasamy, M; Kumar, N; Tang, Z; Mishra, RS; Liaw, PK. Application of machine learning and RSM hybrid models for predictive selection of FSW windows. Comput Ind; 2021; 129, [DOI: https://dx.doi.org/10.1016/j.compind.2021.103456] 103456.

32. Komarasamy, M; Liaw, PK. Multi-criteria optimisation of friction stir welding parameters for EN-AW 6061 using RSM and ANN surrogates. Adv Manufac; 2022; 10, pp. 220-235. [DOI: https://dx.doi.org/10.1007/s40436-022-00376-1]

33. Li, OO et al. The influence of tool pin geometry and speed on the mechanical properties of friction stir processed AA1050. Materials; 2021; 15, 13 4684. [DOI: https://dx.doi.org/10.3390/ma15134684]

34. Liu, H; Chen, W; Zhao, L. Machine learning-assisted multi-response optimization of FSW parameters. Mater Design.; 2024; 231, 112017. [DOI: https://dx.doi.org/10.1016/j.matdes.2023.112017]

35. Luo, Z; Sun, Y; Li, W; He, J; Luo, G; Liu, H. Evolution of microstructures, texture and mechanical properties of Al–Mg–Si–Cu alloy under different welding speeds during friction stir welding. Metals; 2023; 13, 6 1120. [DOI: https://dx.doi.org/10.3390/met13061120]

36. Meyghani, B; Awang, M. The influence of the tool tilt angle on the heat generation and the material behavior in friction stir welding (FSW). Metals; 2022; 12, 11 1837. [DOI: https://dx.doi.org/10.3390/met12111837]

37. Mishra, RS et al. Effect of cone angle of cylindrical pin in double-sided friction stir welding on mechanical performance. J Mater Sci Technol; 2021; 70, pp. 96-107. [DOI: https://dx.doi.org/10.1016/j.jmst.2021.06.001]

38. Mukhopadhyay, T; Banerjee, D; Sen, R. Optimization of FSW of dissimilar aluminum alloys using Grey Relational Analysis and RSM. J Manuf Process; 2022; 75, pp. 540-550. [DOI: https://dx.doi.org/10.1016/j.jmapro.2022.04.018]

39. Newishy, M; Jaskari, M; Järvenpää, A; Fujii, H; Abdel-Aleem, HA. Friction stir welding of dissimilar Al 6061–T6 to AISI 316 stainless steel: microstructure and mechanical properties. Materials; 2023; 16, 11 4085. [DOI: https://dx.doi.org/10.3390/ma16114085]

40. Panda, B; Garg, A; Jian, Z; Heidarzadeh, A; Gao, L. Characterization of the tensile properties of friction stir welded aluminum alloy joints based on axial force, traverse speed, and rotational speed. Front Mech Eng; 2016; 11, 3 pp. 289-298. [DOI: https://dx.doi.org/10.1007/s11465-016-0397-z]

41. Park, SH; Kim, YG; Kang, CY. Optimization of FSW process using Taguchi and RSM approaches. Mater Design.; 2021; 206, 109734. [DOI: https://dx.doi.org/10.1016/j.matdes.2021.109734]

42. Rahimzadeh Ilkhichi, A; Soufi, R; Hussain, G; Vatankhah Barenji, R; Heidarzadeh, A. Establishing mathematical models to predict grain size and hardness of the friction stir-welded AA 7020 aluminum alloy joints. Metall Mater Trans B; 2015; 46, 1 pp. 357-365. [DOI: https://dx.doi.org/10.1007/s11663-014-0205-x]

43. Sabry, I; El-Deeb, MSS. Multi-weld quality optimization of friction stir welding using grey-based Taguchi. Manag Production Eng Rev.; 2024; 15, 2 pp. 45-55.

44. Sabry, I; Gadallah, N; Abu-Okail, M. Optimization of friction stir welding parameters using response surface methodology. IOP Conf Ser Mater Sci Eng; 2020; 973, [DOI: https://dx.doi.org/10.1088/1757-899X/973/1/012017] 12017.

45. Shash, AY; El-Moayed, MH; Abd Rabou, M; El-Sherbiny, MG. A coupled experimental and numerical analysis of AA6063 friction stir welding. Proc Inst Mech Eng C J Mech Eng Sci; 2022; 236, 15 pp. 8392-8400. [DOI: https://dx.doi.org/10.1177/09544062221091416]

46. Singh T, others. Processing of friction stir welded AA6061-T6 joints with nanoparticles (Al2O3 / TiO2): microstructure and mechanical characterization. ---. 2021.

47. Soto-Díaz, R; Vásquez-Carbonell, M; Escorcia-Gutiérrez, JA. A review of artificial intelligence techniques for optimizing friction stir welding. Eng Sci Technol Int J; 2024; 37, pp. 102-126. [DOI: https://dx.doi.org/10.1016/j.jestch.2023.12.005]

48. Suresh S, others. Strategized friction stir welded AA6061-T6/SiC composite lap joint suitable for sheet metal applications. ---. 2022.

49. Tamadon, A; Pons, DJ; Sued, K et al. Thermomechanical grain refinement in AA6082-T6 thin plates under bobbin friction stir welding. Metals; 2018; 8, 6 375. [DOI: https://dx.doi.org/10.3390/met8060375]

50. Tinguery, KMS; Rahem, A; Nadeau, F; Fafard, M. Friction stir welding parameters development of AA6061-T6 extruded alloy using a bobbin tool. Eng Proc; 2023; 43, 50. [DOI: https://dx.doi.org/10.3390/engproc2023043050]

51. Toumi, A; Khalfallah, M; Kharrat, M. Review of bobbin tool friction stir welding: recent advances and applications. J Market Res; 2024; 29, pp. 162-182. [DOI: https://dx.doi.org/10.1016/j.jmrt.2024.01.017]

52. Trueba, L; Torres, MA; Johannes, LB; Rybicki, D. Process optimization in the self-reacting friction stir welding of aluminum 6061–T6. Int J Mater Form; 2017; 11, 4 pp. 559-570. [DOI: https://dx.doi.org/10.1007/s12289-017-1365-4]

53. Vahdati, M; Eslami, A; Maleki, E. Multi-response optimization of AA6061-T6 friction stir welding using RSM. Measurement; 2020; 165, [DOI: https://dx.doi.org/10.1016/j.measurement.2020.108167] 108167.

54. Vairis, A; Li, W; Gao, Y. Investigation on the effects of welding speed on bobbin tool friction stir welding of 2219 aluminum alloy. Metall Mater Trans; 2020; 26, pp. 1830-1840. [DOI: https://dx.doi.org/10.1007/s11663-020-01830-5]

55. Vijayan, S; Raju, R; Rao, SRK. Multiobjective optimization of friction stir welding process parameters on aluminum alloy AA 5083 using Taguchi-based grey relation analysis. Mater Manuf Process; 2010; 25, 11 pp. 1206-1212. [DOI: https://dx.doi.org/10.1080/10426910903536782]

56. Wang, FF; Li, WY; Shen, J et al. Global and local mechanical properties and microstructure of bobbin tool friction-stir-welded Al-Li alloy. Sci Technol Weld Join; 2016; 21, 6 pp. 479-483. [DOI: https://dx.doi.org/10.1080/13621718.2015.1132128]

57. Wang, J; Liu, Q; Xu, R. Hybrid optimization of FSW parameters using RSM and particle swarm optimization. Int J Adv Manuf Technol; 2022; 122, 3–4 pp. 1023-1036. [DOI: https://dx.doi.org/10.1007/s00170-022-09112-9]

58. Yacout, GG; Shash, AY; Hegazi, HA; Elsherbiny, MG. Application of response surface methodology for optimizing friction stir welding parameters of AA6082-T6 aluminium alloy ⋆. J Mater Res Technol; 2025; 37, pp. 3765-3776. [DOI: https://dx.doi.org/10.1016/j.jmrt.2025.07.081]

59. Yang, B; Zhou, C; Li, Y; Wang, H. Microstructural evolution and mechanical properties of AA6061 friction stir welds. J Mater Process Technol; 2018; 252, pp. 680-689. [DOI: https://dx.doi.org/10.1016/j.jmatprotec.2017.10.032]

60. Zhang, J; Huang, Y; Li, W. Multi-objective optimization of AA2024 friction stir welding parameters using RSM and desirability function. J Mater Res Technol; 2020; 9, 6 pp. 13507-13517. [DOI: https://dx.doi.org/10.1016/j.jmrt.2020.09.091]

61. Zhang, L; Ma, X; Wang, Y. Generation of regression models and multi-response optimization of FSW parameters for aluminum alloys using RSM. Can J Mech Eng; 2019; 47, 4 pp. 456-472. [DOI: https://dx.doi.org/10.1139/tcsme-2019-0162]

62. Zhang, Y; Wu, D; Sun, H. Numerical simulation and defect prediction in bobbin tool friction stir welding. Comput Mater Sci; 2021; 188, [DOI: https://dx.doi.org/10.1016/j.commatsci.2020.110212] 110212.