Investigating the effects of ultrasonic assistance on TIG welding of AA7075 alloys: a machine learning-based optimization study using RSM-PSO

Abstract This study explores the impact of ultrasonic assistance on TIG welding of AA7075 alloys, leveraging a machine learning-based optimization approach combining Response Surface Methodology (RSM) and Particle Swarm Optimization (PSO). A central composite design matrix was developed to investigate the effects of process parameters on microhardness and weld defects. A predictive model was constructed, utilizing process parameters as inputs and microhardness as the output. PSO optimization was then applied, followed by experimental validation. The model demonstrated high accuracy, with R-squared values of 0.9808 and 0.9862 for conventional and ultrasonic-assisted TIG welding. Confirmation tests showed an error margin of less than 1%. The optimal process parameters under ultrasonic vibration were identified as welding current (50.38 A), gas flow rate (12.42 l min −1 ), and filler material (ER5356). The predicted microhardness (153.16 HV) closely matched the actual value (150.71 HV), with an error of 1.6%. Tensile and fractography analyses further validated the optimized welding parameters. This research showcases the potential of integrating ultrasonic vibration with RSM-PSO optimization to enhance weld quality and mechanical properties of AA7075 alloy joints, offering valuable insights for industrial applications.