Experimental and data-driven optimization of tool geometry controlled mechanical performance of friction stir welded AA6061-T6 joints

Abstract Friction stir welding (FSW) of Al6061-T6 requires careful control of heat input and material flow because weld quality depends strongly on tool geometry, spindle speed, and feed rate. This study evaluates the combined effects of pin profile (cylindrical, square, and triangular), spindle speed (800, 1000, and 1200 rpm), and feed rate (6, 8, and 10 mm/min) on the tensile strength, impact strength, and microhardness of Al6061-T6 joints produced using H12 die steel tools. A full factorial experimental design was employed, followed by analysis of variance, artificial neural network (ANN) modelling, and multi-objective optimization using NSGA-II. The results show that tool geometry was the dominant factor, with the square pin producing superior joint performance because its larger effective contact perimeter promoted more uniform stirring, improved consolidation, and reduced pore/void formation. Tensile and impact performance were maximized at the square pin, 800 rpm, and 8 mm/min, whereas the highest microhardness was obtained at the square pin, 1200 rpm, and 6 mm/min. The ANN model predicted the responses with high accuracy (overall R ≈ 0.988), and NSGA-II yielded improvements of 27.66% in tensile strength, 46.56% in impact strength, and 13.68% in microhardness over unoptimized conditions.