PREDICATION OF WELDBEAD GEOMETRY USING ARTIFICIAL NEURAL NETWORK OF AL 6061 ALLOY

Abstract

MIG has been a functional welding tool due to accuracy for joining metal parts. In any industrial applications, a good quality weld bead geometry is required. In the present work, Taguchi technique was adopted to carry out experiments and established an interrelationship between input parameters and weld bead geometry through an artificial neural network (ANN) for Al6061 alloy. The Levenberg-Marquardt algorithm was used to model for back propagation and the adequacy of modeled neural network is checked by validating data from trained data. The predictions of the model of artificial neural network obtained are closer to experimental result, which reveals that modeled neural network is feasible means for predicting weld bead geometry.