Using Hybrid Artificial Neural Network and Particle Swarm Optimization Algorithm for Modeling and Optimization of Welding Process

Abstract

This study addresses a hybrid procedure used for modeling and optimization of gas tungsten arc (GTA) welding process of AL5052 alloy. There are different process input parameters among which welding current ([Formula: see text], frequency ([Formula: see text], welding speed ([Formula: see text], and gap ([Formula: see text] are the most important ones considered in GTA welding process. Furthermore, heat affected zone (HAZ) is considered as the most important quality measure of the process. To gather the required data for the modeling and optimization purpose, design of experiments (DOE) approach has been used. Image processing technique is used to take accurate measurements of HAZ values. In order to determine the relationship between process input variables and output measures, artificial neural networks (ANNs) have been used. Then, the trained ANNs have been used to find the optimal value of the outputs using particle swarm optimization (PSO) algorithm. Experiments have been done to verify the optimal levels of the input parameters. Verification results demonstrate that the proposed ANN-PSO procedure is quite efficient in modeling and optimization (with about 4% error) of GTA welding process.