Investigation of the optimal parameters for butt joints in a friction stir welding (FSW) process with dissimilar aluminium alloys

Aluminum alloys are used in the modern automotive industry because they are lightweight. However, it is establised that conventional fusion welding processes affect weld performance. In this study, friction stir welding (FSW), also known as solid-state welding, was used to weld dissimilar aluminum alloys, AA6061-T6 and AA5083 plates. Response surface methodology based on Box-Behnken design (BBD) was used to investigate the appropriate process parameters. In this study, the effects of rotation speed (S), welding feed rate (f), and work angle (θ) were investigated. These three factors were examined at three levels over 17 experimental runs. The design was used to conduct experiments and develop mathematical regression models. Variance analysis was performed to examine the adequacy of the developed models. Finally, the effects of the process parameters on the mechanical properties of welded alloyes were investigated using mathematical models based on the macrostructure, microstructure, chemical composition, and fracture characteristics of the joints using SEM. The investigation found that the optimum welding parameters are a rotational speed of 777 rpm, welding speed of 44 mm min−1, and a work angle of 0.75°. Furthermore, the results confirm that the mathematical models and experiments are consistent.


Introduction
Welding of dissimilar aluminum alloys (Series 5000: Al-Mg and 6000: Al-Mg-Si) by fusion welding is very challenging because brittle intermetallic compounds (IMCs) often form from eutectic reactions during the solidification process, leading to hot cracking [1]. Weld cracking caused by solidification is highly sensitive and often dependent on the chemical composition of the workpiece and filler metal, as well as the degree of dilution. These parameters more often lead to failure of dissimilar welds compared to fusion welding. An aluminium oxide film has a melting point of approximately 2,040°C, which is higher than aluminum, with a melting point of approximately 680°C. Therefore, oxides must be entirely removed prior to welding. If these oxides are not removed, welding will require high heat to melt the oxide film. When aluminum oxide on the surface is annealed, the aluminum metal often evaporates to form a vapor. Welding aluminum penetrates the surface of a metal workpiece when heated [2]. Additionally, aluminum expands twice as much as steel owing to the heat. Aluminium is easily deformed when heated by welding [2]. The thermal conductivity of aluminum is approximately six times higher than that of steel. Therefore, the heat required to weld aluminum plates must be more highly concentrated in an aluminum plate than when welding steel plates. Welding thick workpieces will lead to a lack of fusion in the weld [2]. With complex dissimilar metal welding, it is necessary to avoid melting during the fusion welding process. This can be done by applying solid-state welding using a friction stir welding process. Solid-state welding is suitable for welding various aluminum alloys to address the aforementioned problems. Friction stir welding does not cause the weld to form molten pools. It only causes agitation under plastic deformation [3][4][5][6]. Numerous studies have reported that friction stir welds have high weld efficiency [7][8][9][10].
AA6xxx and AA5xxx aluminum alloys are high-strength and are formable. They are widely used the automotive, aerospace, marine, and transportation industries. Earlier studies, such as that of Elengovan et al [11] examined the effects of tool characteristics and FSW welding variables on the tensile strength of aluminum AA6061 using finite element simulations. The flow of the metal and plastic deformation in the weld area were higher than for other shapes. Rajakumar et al [12] studied the effects of FSW on the tensile strength, hardness, and corrosion resistance of AA6061 aluminum alloy welds. They found that the rotation speed and tool shoulder diameter influenced the mechanical properties and corrosion resistance of FSW welds. Lee et al [13] studied parameters in friction welding that significantly affected the weld microstructure and variation of its mechanical properties. Ramalu et al [14] studied the effects of rotational speed, travel speed, plunge depth, and shoulder diameter on weld properties. They found that increased rotational speed and plunge depth reduced travel speed and increased weld heat while reducing weld defects in an AA5xxx aluminum alloy. Kwon et al [15] examined Al5052 aluminum alloy joints formed with an SFW process. Investigation of rotation speeds of 500−3000 rpm was done under a constant traverse speed of 100 mm min −1 . No defects were found in the onion ring microstructure for all the operational factors. It was also found that as rotation speed increased, the size of the onion ring microstructure was enlarged. It was found that at higher rotation speeds, joint hardness was greater than that of the base metal. The strength was close to that of the base metal, while joint elongation was lower than the base metal. Han et al [16] investigated the effects of rotational and travel speeds on the mechanical properties of welds. It was found that higher travel and rotational speeds decreased weld strength, while weld elongation increased. Moshwan et al [17] used the FSW process while examining the mechanical properties and microstructure of 3 mm thick AA5052-O aluminum alloys. Joints were produced with a constant travel speed of 120 mm min −1 and rotation speeds of 800 -3000 rpm. It was found that at a rotation speed of 1000 rpm, a maximum tensile strength of 132 MPa was achieved, which was 74% of the strength of the base material. From EDS analysis, intermetallic Mg2Al3 compounds were found in the stirring zone.
Several studies related to dissimilar welding of AA 5xxx and AA 6xxx grade aluminum alloys, such as that of Kasman [18], investigated the optimal parameters for FSW welding using the Taguchi method and grey relational analysis. It was found that the tool shoulder diameter, tool diameter, welding speed, and tool rotation speed significantly affect weld efficiency. Leitão et al [19] studied FSW welding of AA5083 aluminum alloy with AA6082 to examine material flow. It was found that the AA6082 aluminum alloy exhibited higher weldability due to its better flow during plastic deformation at high temperatures than the AA5083 alloy. Palanivel et al [20] studied the effects of FSW welding parameters, including rotation speed and tool profile, on the microstructure and tensile strength of AA5083-H111 and AA6351-T6 aluminum alloys. Alvarez et al [21] conducted experiments on the FSW welding process with a right-hand-threaded cylinder on AA6082-T6 to AA5754-H111aluminum alloy plates. Jamshidi-Aval et al [22] studied FSW welding temperatures using a 3D model with the ABAQUS program for AA6061-T6 and AA5086-O aluminum alloys.
Some studies suggested experimental design approaches for predicting and optimizing FSW welding to achieve desired mechanical properties. For example, Nakowong and Sillapasa [23] studied optimization of FSW welds using the Taguchi method and analyzed the variance in tensile strength and weld hardness. Zamani et al [24] studied welded Al-SiC aluminum alloys using an FSW process with response surface methodology (RSM) to determine optimal weld parameters. Sefat et al [25] experimentally welded AA5052 H18 aluminum alloys using the FSW process. Their experimental design was performed using RSM to determine the optimal welding parameters. Naqibi et al [26] studied the welding of copper-aluminum tubes using the FSW process. These factors were optimized using Box-Behnken design (BBD) in conjunction with the RSM method. Kumar and Sharma [27] experimented with welding dissimilar joints of AA5083-O and AA6082-T6 aluminum alloys of tailor-welded blanks using RSM to determine the factors involved in FSW welding. Chakradhar and Narendranath [28] designed an experiment using the RSM method for joining an AA6061 aluminum alloy using the FSW process to determine the optimal parameters for tensile strength and weld hardness. Mirabzadeh et al [29] examined heat generated by FSW welding on polypropylene sheets through process factor analysis using the RSM method with Box Behnken design. Ahmadnia et al [30] studied the effects of FSW welding parameters on the tensile strength, hardness, and elongation of AA6061 and AA5051 aluminum alloy welds using the RSM method. Many additional cases of experimental design for predicting and optimizing FSW welding have been presented in the literature. Previous research suggests that an appropriate experimental design can effectively predict the experimental outcomes of FSW welding.
To the best of our knowledge, there have been no reports of the FSW welding process used for welding AA6061-T6 and AA5083 aluminum alloys. Furthermore, the morphology of these welds has not been analyzed. The current research aims to design a statistical experiment and apply response surface methodology (RSM) using a quadratic model that employs BBD. Furthermore, to optimize welding parameters, the effects of the rotation speed (S), travel speed ( f ), and work angle (θ) on the tensile strength and hardness of the weld were investigated. Additionally, research has shown that an application method can be used to select the optimum operating conditions to achieve the desired weld properties. Finally, fracture analysis of the welds was conducted to confirm the response.

Materials and experimental equipment
The chemical composition and mechanical properties of AA6061-T6 and AA5083 aluminum alloys are shown in tables 1 and 2, respectively. These alloys were experimentally employed with dimensions 50 × 175 mm (width x length) and a 6 mm thickness. They were cut by sawing and then surface trimmed using a milling machine, as depicted in figure 1(a). The welding tool was a tapered threaded cylindrical shape. Its dimensions and geometry are shown in figure 1(b). Welding tools were fabricated using SKD-11 alloy tool steel, hardened at 1050°C, and then oil cooled. The workpiece clamping process for welding is shown in figure 1(c). The FSW process used a computer numerical control (CNC) machine to maintain the desired welding parameters.
In our experiment, an AA5083 aluminium alloy is the advancing side, and an AA6061-T6 aluminium alloy is the retreating side, since the advancing side has a higher plastic flow than the retreating side. [23] Additionally, the AA5083 aluminium alloy has a lower hardness than AA6061-T6, resulting in higher flow.

Mechanical testing
FSW welded specimens were cut to prepare tensile and hardness test specimens, as shown in figure 2. To prepare the tensile specimens, a VMC MACHINE CYCLONE-610 was used to reduce the width of the specimen, while maintaining their thickness and surface features. Tensile tests were conducted using a universal testing machine (Model: HD B616-2-60T, at Nakon Phanom University, Thailand) according to the American Society for Testing of Materials, Standard ASTM E8M [31], as shown in figure 2(a). The strain rate was 15 s −1 in the tensile test of this experiment.
A welding hardness test was performed across the SZ, HAZ, TMZ, and BM using a Vickers micro-hardness tester (IQUALITROL Model, MHV-2000Z Series, Dongguan, China) at a load of 9.81 kgf, indentation distance of 0.2 mm, and indentation time of 10 s. The distance between the test points was 15 mm from the center of the workpiece on both the advancing side (AS) and retreating side (RS), as shown in figure 2(b). They were then polished using different grades of emery paper, from P180-P800. The samples were then etched for 25 s using a mixture of 100 ml of H 2 O and 3 ml of HF. Finally, the samples were rinsed with distilled water and clean with alcohol. A hot-air gun was used to blow dry the samples. Investigation of the chemical composition of the joint was done using scanning electron microscopy (SEM).

Experimental design
The optimal factors for the welded joint tensile strength and hardness were determined using a Box-Behnken design (BBD). The factors examined in the study were the rotational speed (S; X 1 ), welding feed rate ( f; X 2 ), and  work angle (θ; X 3 ). Factor levels were determined based on relevant research and limitations of the experimental tools. The levels of the factors were low (−1), medium (0), and high (1), as shown in table 3. The experiment and design used Design-Expert and MINITAB 19 statistical software packages. Design-Expert software was used for graphical optimization and verification of the model coefficient of determination (R 2 ). Analysis of variance (ANOVA) was performed using MINITAB. The response equation is shown as equation (1), where Y is the response and composite desirability (D). Composite desirability of the result is between 0-1. If D is equal to 1, the result has a complete composite desirability.  Table 5 and 6 show analysis of the relationships between the response factors for the ultimate tensile strength (Ts) and weld hardness (HV).

Development of the RSM for mechanical properties
The response of the process factors related to process quality was determined. Regression analysis was employed to develop a mathematical model using quadratic polynomials for Ts and HV. In model development, statistical analysis was used to assess the validity of the full quadratic models using analysis of variance (ANOVA) and the coefficient of determination (R 2 ). Mathematical models using a polynomial regression of responses, linear terms, quadratic equations, and interaction terms are demonstrated by equations (2) and (3). ( )  where S is the rotational speed, f is the weld feed rate, and θ is the work angle. Ts and HV respectively represent the tensile strength and hardness of a weld. ANOVA was performed to verify the precision of the developed mathematical models. These results are presented in table 5. For the tensile strength values presented in this table, the model p-value for Ts is less than 0.05, indicating that its conditions are significant for the terms S, f, S 2 , f 2 , and θ 2 , which have p-values equal to 0.011, 0.019, 0.000, 0.044, and 0.006, respectively. In contrast, θ, Sf, Sθ, and fθ had p-values greater than 0.05, indicating that the model conditions did not significantly impact Ts. However, for 'Lack-of-Fit' in table 5, the p-value is higher than the critical value of 0.309, which is greater than 0.05, so the null hypothesis cannot be rejected, indicating that the model is significant. Therefore, the regression model is suitable for predicting the Ts values of the weld. Furthermore, considering that the R 2 value of the model approaches a value of 1, the model can be used to establish predictive equations.   Table 6 presents the ANOVA results. The HV data show that the model p-value for the HV is less than 0.05, indicating that the model conditions are significant. Therefore, terms S 2 and Sθ are 0.003 and 0.027, significantly impact HV. However, the other terms in this table have p-values higher than 0.05, indicating that these model conditions did not significantly impact HV. However, the 'Lack-of-Fit' in table 6 has a p-value higher than the critical value of 0.128, which is greater than 0.05, indicating that the model has good fit. Therefore, it can be concluded that the regression model can be suitably used to predict the HV values of the weld. Furthermore, considering that the R 2 value of the model approaches 1, the model can be used to establish the predictive equations.

Optimization analysis
Multiple response optimization was done in the study of impacts of welding AA5083 and AA6061-T6 aluminum alloys using the FSW welding process on the Ts and HV values of the weld. These results are presented in table 7. It was found that the optimal conditions in the experiment were a rotational speed of 777.77 rpm, a welding feed of 44.24 mm min −1 , and a work angle of 0.75°, which resulted in an average hardness of 80.06 HV and average Ts of 190.06 MPa, as shown in figure 3. These optimal results for FSW welding were used for numerical optimization in MINITAB. The composite desirability (D) of this investigation was 0.9058.

Confirmation analysis
Optimal FSW conditions were a rotation speed of 777.77 rpm, welding feed rate of 44.24 mm min −1 , and work angle of 0.75°. The experiment was repeated to confirm the results. In the confirmation analysis, the rotation speed was adjusted to 777 rpm, the welding feed rate to 44 mm min −1 , and the work angle to 0.75°. Three workpieces were welded, and the Ts and HV values were investigated to determine the response. Table 8 shows the results of the confirmatory experiments with the statistical analysis. It was found that the experimental confirmation of the mean Ts value was 185.42 MPa, the error was 2.44%, and the experimental confirmation of the mean HV value was 76.62 HV, an error of 4.28%. The response from the confirmation experiment was within the acceptable range, i.e., the total error of all responses was not more than 5%.

Analyzing mechanical properties
The response surface of the maximum Ts was established from equation (2) to predict Ts. It was plotted and shown in figure 4. It illustrates perturbation and 3D response surface graphs. Figure 4(a) shows the effect of the operational parameters (A: rotational speed, B: welding feed rate, and C: work angle) on the Ts values of the joints. It was observed that the impact rotational speed on Ts was the greatest followed by the welding feed rate and work angle. Figures 4(b)-(d) illustrate the parameters of the response surface curve to the joint Ts values. Intermediate parameter levels yielded maximal Ts values. However, it was found that when the parameter levels increased or decreased, the weld Ts value decreased. The contour plot shows that the relationship between the process parameters and Ts exhibits a nonlinear response. The curve in the center shows the maximum Ts, whereas slight changes result in reduced Ts values. Figure 5 illustrates perturbation and 3D response surface graphs. Figure 5(a) shows the effect of the parameters (A: rotational speed, B: welding feed rate, and C: work angle) on the HV of the joints. It was observed that the rotational speed, welding feed rate, and work angle did not significantly impact the HV values of the joints. Figures 5(b)-(d) illustrate the parameters affecting the response surface curve upon joint hardness values. Intermediate parameter levels produced maximal HV values. When the parameter levels were increased or decreased, the contour plot shows that the relationship between the process parameters and HV has a nonlinear effect. The center of the curve shows a maximal HV, whereas slightly changed parameters reduced the HV values.

Macrostructure analysis
Macrostructural analysis was performed to determine the FSW characteristics of AA6061-T6 and AA5083 aluminum alloy joints under ideal experimental conditions, i.e., a rotational speed of 777 rpm, a welding speed of 44 mm min −1 , and a work angle of 0.75 degrees. A flash defect was found at the beginning of welding that disappeared as the welding distance increased. The welded surface exhibited no defects, as shown in figure 6. Additionally, when examining the weld cross-section, a tunnel defect was found at the bottom of the weld owing to insufficient metal flow, as shown in figure 7. The tunnel defect resulted from a reduction in fluidity and flow of material to the end of the tool on the advancing side. In this study, tilting the specimen reduced tunnel defects [32].

Microstructural and chemical analyses
Investigation of the microstructure of the joints was performed at a rotational speed of 777 rpm, a welding feed rate of 44 mm min −1 , and a working angle of 0.75°, which were previously determined as the optimal welding parameters. Figure 8(a) shows that the TMAZ zone of the retreating side had a long grain structure caused by rotation of the tool inducing material flow. The TMAZ zone of the retreating side is not homogeneous, thus on the joint strength. In figure 8(b), the TMAZ zone of the advancing side has a fine grain structure, and the material is homogeneous with no defects. Since the advancing side is hotter than the retreating side, sufficient heat dissipation and plastic deformation made the material flow uniform in the stirring zone, as shown figure 8(c). The SZ zone has a long, layered structure with material flow at the advancing and retreating sides, resulting in the SZ zone being homogeneous. However, a tunnel defect formed at the bottom of the joint.
EDS analysis of AA6061-T6 and AA5083 aluminum alloy joints of the FSW process is shown in figure 9. Mapping analysis was employed. The chemical composition of the weld zone was primarily Al, Mg, Si, F, and O, as shown in table 9. Observations indicate that the chemical composition of the joints consists of an Mg solid solution in Al and Mg 2 Al 3 [33] intermetallic compounds, which are the base chemical components of welding materials. Similarly, the investigation found O in the joint, which may have been caused by the process because there is no gas shielding in FSW. When the joint is heated during welding, an Al 2 O 3 compound [23] may have been formed in the weld. However, the amount Al 2 O 3 formed is insignificant and does not impact the mechanical properties of the weld. Figure 10(a) shows the morphology of the surface characteristics at the upper position of the weld. The Al 3 Mg 2 intermetallic compound is spherical and uneven, and Al 3 O 2 has small (dark black) particles adhering on the joint surfaces in figure 10(b). The surface morphology at the center of the weld shows an Al 3 Mg 2 compound. The particles were more uneven here than in the upper region, and minor Al 3 O 2 compounds were distributed on the weld surfaces, as shown in figure 10(c). The surface morphology of the welded bottom is characterized by a non-uniform distribution of Al 3 Mg 2 particles and a small amount of Al 3 O 2 on the weld surface in figure 10(d). Surface morphology at the TMZ location, retreating side, non-uniform distribution, and high and low particle formation are seen on the weld surface. Finally, in figure 10(e), the surface morphology at the TMZ advancing side site exhibits irregular and heterogeneous particle characteristics.

Fracture analysis
The fracture mechanisms of the specimens in this experiment are similar. Joint failure is expected in the TMAZ retreating zone. This is because the retreating side was fabricated using an AA5083 alloy that has poorer mechanical properties than the advancing AA6061-T6 alloy side. Similarly, the bottom of the joint has a defect resulting in rapid fracture due to stress concentration at the defect corner. This suggests that a defect determines the fracture position, endangering the ductile region and TS of the joint. In the current study, fracture of the specimen was investigated for specimens fabricated at a rotation speed of 777 rpm, welding speed of 44 mm min −1 , and work angle of 0.75°. These are the optimal welding parameters in the current experiments. The fracture investigation examined both the retreating and advancing sides. Fracture characteristics are shown in Figure 11. Three fracture zones were examined. They are the top zone (I), the intermediate zone (II), and the bottom zone (III). First, the bottom fracture (III) is damaged near the tunnel defect. A cleavage fracture was observed, and the fractured surface resembles a collection of rocks. This is due to the rapid fracture due to stress concentration at the angle zone of the defect [34]. Furthermore, the fracture surface is perpendicular to stress in the tensile test. Also, it indicates that this area is the crack initiation of the fracture. Subsequently, intermediate   zone (II) fractures show a change from a cleavage fracture to dimpled surface [35]. Intermediate zone damage on both the retreating and advancing sides was similar. This is because the intermediate zone of the joint has excellent mixing of the welding materials and no defects are found in this zone. The results is an increased tensile strength in the intermediate zone as well as a decreased fracture sensitivity resulting in plastic deformation in this area. Finally, the top zone (I) of the joint has a fracture surface that is a combination of a dimple and cleavage fracture. However, the size of the dimple and cleavage are smaller. Top zone damage on both the retreating and advancing sides is similar. A large amount of phase precipitation was also found at the fracture surface.

Discussion
In this study, optimization of dissimilar aluminum joints between FSW AA6061-T6 and AA5083 alloys was investigated. Our experimental design used a response surface method (RSM) with Box-Behnken design (BBD).
The experimental parameters were rotational speed (600 800 rpm), welding feed rate (20-40 mm min −1 ), and work angle (0-1.5°). Five median replicate experiments were performed for a total of 17 runs. Then, the results were analyzed for prediction of Ts and Hv as a function of the independent variables (X 1 , X 2 , and X 3 , as defined in section 2.3) using a quadratic equation. Investigation of the experimental factors showed that rotational speed and welding feed rate significantly influenced Ts, but the work angle is not. It was found that no factor significantly impacted the hardness of the joints. Then, the relationship of factor responses was analyzed using a regression model at a confidence level of 0.05 (α = 0.05). It was found that the model for Ts had an R 2 = 0.9635  and 0.8388 for HV. Lack of fit p-values were 0.309 and 0.128 for Ts and Hv, respectively. These are higher than 0.05 [21], indicating that the models can be used to predict Ts and HV as equations (2) and ( 10. The results are different from previous work, as earlier studies focused on the mechanical properties of the joints. Additionally, most intermetallic compounds and fracture characteristics of the joints have not been studied. Therefore, this study investigates the mechanical properties, intermetallic compounds and fracture characteristics of dissimilar AA5083 and AA6061-T6 aluminum welds with the expectation that this research will be helpful to those interested in further studies.

Conclusion
The optimal parameters of the FSW process for Al6061-T6 and AA5083 aluminum alloys were determined in the current study. Additionally, the effects rotation speed, welding feed rate, and work angle on the Ts and HV Figure 11. Fracture analyzing dissimilar welding of AA5083 and AA6061-T6 aluminum alloys. − Mathematical modeling with BBD based on analysis of variance revealed that the terms, S, f, S 2 , f 2 , and θ 2 , significantly impact tensile strength. Additionally, θ, Sf, Sθ, and fθ did not significantly affect the Ts regression model. An investigation of hardness showed that the S 2 and Sθ terms significantly affected HV, whereas other terms did not.
− Simultaneous optimization of mechanical properties was done using a desirability function and numerical approach. These results suggest a rotation speed of approximately 777 rpm, a welding speed of approximately 44 mm min −1 , and a work angle of approximately 0.75°as an optimal solution that produces a 190.06 MPa tensile strength and an 80.05 HV joint hardness.
− Verification of the obtained optimal results through a confirmatory experiment and the experimental findings showed that the proposed approach could predict the optimal solutions with overall error values lower than 5%.
− The microstructure of the joints was investigated in the SZ area, where the material flowed well and was consolidated with long grain layers. The TMAZ region of the retreating side had a long grain structure caused by rotation of the pin, which induced the material to flow. The structure in this region is inconsistent, thus the strength of the joint corresponds to the fracture position of the tensile specimen . Finally, the TMAZ region of the advancing side of the structure showed fine grains and microstructural harmony. No defects were found. Since the advancing side is hotter than the retreating side, the heat is sufficient and plastic deformation leads to uniform material flow.
− EDS investigation revealed that Al, Mg, Si, F and O were centrally mixed. Consequently, an Al 3 Mg 2 intermetallic compound [32] was formed, and an Al 3 O 2 compound (dark black) was found, characterized as particles on the weld surfaces. The particle characteristics are discontinuous and inhomogeneous at the TMZ position on the retreating side. Similarly, the morphology of the TMZ position on the advancing side is uneven and heterogeneous.