Mathematical modeling of the thermal efficiency of electric arc welding using artificial neural networks

Abstract

This paper deals with the problem of the estimate of the arc welding thermal efficiency, especially in relation to the Gas Metal Arc Welding process. Despite being a parameter of great importance, the efficiency is determinate, oftentimes, from tabulated values given by the technical standards, which can affect the metallurgical analysis, since these values not always can consider each possible variation in the welding parameters. In this context, this paper proposes the development of a mathematical model through Artificial Neural Networks (ANNs) to be used in conjunction with a continuous flow calorimeter, in order to obtain a preliminary estimate for efficiency and avoid expenses with experimental tests that would produce suboptimal results. The experimental data were obtained through tests of the welding process in the company Bruning Tecnometal Ltda, and the simulations were carried out using the computational tool Matlab. The proposed model was validated using the k-fold cross-validation method and presented a low mean relative error.