Apply for Back-Propagation Neural Network to Control a GMA Welding Process

Recently, GMA (Gas Metal Arc) welding process has been automated in an attempt to gain high efficiency, high productivity and low costs. With the combination of sensors and mathematical models, increased effectiveness in control of the robotic welding process was achieved. As a result of these short-comings, much research and development works have been concentrated on sensing and control methods to enhance the robotic arc welding. This paper presents a new Genetic Algorithm (GA) to select the optimal architecture of the back-propagation neural network and compared with that of engineer's experience. It is shown that learning approach with the optimal structure of back-propagation neural network could be applied to predict the bead geometry such as bead width and bead height in robotic GMA welding process and compared between the bead geometry by calculated from the developed model and experimental results.