Genetic-neural optimization approach for gas metal arc welding of dissimilar aluminium alloys of AA5083-O/AA6061-T6

Abstract

Distinct alloys have different chemical and thermal characteristics, making it difficult to weld together. For the purpose of maximizing tensile strength and weld hardness of the joint, the gas metal arc (GMA) welding process for the dissimilar aluminium alloys AA5083-O and AA6061-T6 was modelled and optimized in the current study. A genetic-neural approach was attempted, in which optimal artificial neural network (ANN) was applied to model the process and genetic algorithm (GA) approach was extended to optimize the parameters. The proposed genetic-neural (GA-ANN) approach was also compared to the traditional response surface methodology (RSM). In predicting the reactions of a GMA welded joint made of two different alloys, AA5083-O and AA6061-T6, the suggested optimum ANN model was shown to be more accurate (error 6%). The genetic-neural optimization technique has less inaccuracy (approximately 5% error) than the RSM optimization approach, however the more computational time was required to select GA-ANN parameters.