EXPERIMENTAL AND OPTIMIZATION-BASED ANALYSIS TO MAXIMIZE MECHANICAL-RELATED ATTRIBUTES OF FRICTION STIR WELDED CDA101 CU ALLOY JOINTS

This paper deals with the identification of optimized process parameters for FSW of 6[Formula: see text]mm thick CDA 101 Cu alloy plates. Three distinctive parameters, the tool’s traverse speed, rotational speed, axial load, have been considered as self-reliant variables for formulating a numerical model based on nonlinear regression-type approach, with the objective of anticipating mechanical attributes of joints. From the surface- and contour-type plots, it was revealed that the axial load in combination with rotational speed has played a dominant role in influencing the tensile and yield strength of the CDA101 Cu alloy joints. At the same time, the traverse speed of the tool has played a dominant role in enhancing the ductility of the CDA101 joints through the generation of ample volumes of frictional heat, which in turn have facilitated both coarsening of grain structures and softening of materials in an ideal manner. Employing tool rotational speed, traverse speed and axial load in the ranges of 1750–1900[Formula: see text]rpm, 23–27[Formula: see text]mm/min and 5.75–6.5[Formula: see text]kN, respectively, was found to enhance the tensile strength of the CDA101 Cu alloy joints. FSW process parameters attained through the numerical meta-model of RSM have led to the attainment of the highest value of tensile strength of 209.5[Formula: see text]MPa together with a 141.4[Formula: see text]MPa yield strength with very minimal fitting errors of 0.28% and 0.52%, respectively. A collection of additional experimental data has been employed to verify the formulated empirical-based formula and additionally, an entirely different optimization technique namely PSO (i.e. particle swarm optimization) was implemented for comparison and validation purposes. The formulated meta-model of RSM generated the largest value of % of elongation of 14.23%, though the relative percent error was around 1.988%, which was slightly larger than that of the value (i.e. 14.11%) generated by PSO, with a relative error of 0.58%.