ARTIFICIAL NEURAL NETWORKS FOR PREDICTION OF DEPTH OF PENETRATION IN MIG WELDING.

ARTIFICIAL NEURAL NETWORKS FOR PREDICTION OF DEPTH OF PENETRATION IN MIG WELDING.

Abstract

Depth of penetration is one of important physical characteristic of a weldment. This paper was done on the basis of prediction that there is a relationship between welding parameters and depth of penetration in gas metal arc welding. Dimensionless model and artificial neural network were used as methods for predicting the depth of penetration. The dimensionless model and the artificial neural network were formed, and the analytical expression for depth of penetration was taken from the research done by P.E Murray and A.Scotti for dimensionless model, and the training data or test data which were used in the formation process of the artificial neural network, were used to perform the prediction of the depth of penetration was taken from experiment done by above researchers. Back-propagation neural networks are used to associate the welding process variables with the features of the penetration. These networks have achieved good agreement with the training data and have yielded satisfactory generalization.
Therefore it is concluded that the error rate predicted by the artificial neural network was smaller than that predicted by the dimensionless model, in terms of the depth of the penetration.