Determining Optimum Butt-Welding Parameters of 304 Stainless-Steel Plates Using Finite Element, Particle Swarm and Artificial Neural Network

Abstract

Residual tensile stresses generated during butt welding of plates using arc welding process lead to deformation and deterioration of fatigue strength of welded parts. This research implemented particle swarm optimization (PSO) algorithm to present optimum welding parameters to minimize the tensile residual stresses of butt-welded 304 stainless-steel plates with 4–15 mm thicknesses. A set of 32 experiments was designed using Taguchi method and simulated using ABAQUS commercial software based on element birth-and-death finite element technique. An artificial neural network and PSO were utilized to discover the optimum welding settings. To ensure the accuracy of simulation results, slitting method was implemented to measure residual stresses utilizing digital image correlation technique beside the strain gauges.