Process parameter optimization of submerged arc welding on mild steel AISI 1020 using an artificial neural network trained with multi-objective Jaya algorithm

Submerged arc welding (SAW), renowned for its high deposition rate and superior weld quality, is the go-to method for joining thick metals in heavy structures. However, industry beams and columns welded with SAW can exhibit detrimental defects like undercut, porosity, and burn-through, significantly impacting weld properties. This study addresses this challenge by presenting a multi-objective optimization approach for SAW parameters on AISI 1020 mild steel. Aiming to optimize tensile strength, hardness, and bead width, the study employs Taguchi's design of experiments and couples the multi-objective Jaya algorithm with an artificial neural network (ANN). This synergistic combination yielded optimal process parameters: 417 A welding current, 20.7 mm electrode stick-out, 33.7 V voltage, and 505.8 mm/min transverse speed. These settings translated into exceptional weld characteristics, with ultimate tensile strength reaching 427 MPa, hardness of 73.9 HRB, and bead width of 14.29 mm. Confirmation tests further validated these findings, demonstrating minimal error and solidifying the effectiveness of the optimization approach. This research paves the way for enhanced weld quality and process control in heavy structural applications.