Influence of Shielded Metal Arc Welding Parameters on the Bead Geometry,Based on Artificial Neural Network

This paper presents an artificial neural network simulation, based on 216 set of experimental data for prediction of SMAW bead geometry. Input parameters consist of electrode type, current, travel speed, arc length, transverse motion, and angle of the electrode with respect to the vertical plane along the weld line. outputs of the neural network were bead width, bead height, penetration, and the rate of electrode consumption. The Error analysis in the neural network was performed by well-known back-propagation error reduction method containing two hidden layers. The results showed that the results were basically in good agreements with the available experimental data,indicating a suitable technique for simulation of the SMAW. It is concluded that the established neural network system, in conjunction with back-propagation analysis, is able to determine the bead geometry for a set of specified input parameters, facilitating the design of special purpose arc welds.