An artificial neural network model for multi-pass tool pin varying FSW of AA5086-H34 plates reinforced with Al2O3 nanoparticles and optimization for tool design insight

Abstract This study is a novel attempt to investigate the powder-assisted multi-pass Friction Stir Welding (FSW) of AA5086-H34 joints with different pin geometries, alongside the analysis of the microstructure and mechanical properties. Towards that end, through the Design of Experiments (DOE), five unlike pin geometries were utilized to fabricate 140 joints with different process parameters. An FSW model was also produced by training an Artificial Neural Network (ANN) that correlates the joint properties and FSW parameters. This model was then used as an objective function to optimize the process parameter values in the pursuit of finding the best process parameters and tool pin geometry. Moreover, to investigate the mechanical properties and microstructures of the welds, optical and scanning electron microscopies were employed. With the multi-pass technique, no accumulation of the reinforcing particles was witnessed in the nugget of the joints. The square pin-profiled tool (SQ) generated a joint with the most homogenous material distribution. The SQ tool contributed to the highest Ultimate Tensile Strength (UTS) of 303 MPa (which is about 98% of the UTS value measured for the base metal). The optimization results validated the superiority of the square geometry in achieving the highest UTS and proposed the best values for the geometric parameters of the FSW tool pin.