Study on the Technique of Neural Network and Fuzzy Control for GTAW

An intelligent system including both neural network and fuzzy controller for the gas tungsten arc welding(GTAW) was presented in this paper.The discussion was mainly focused on the application of neural network and fuzzy logic in modeling and controlling the penetration depth as well as the seam tracking.A visual sensor CCD was used to obtain the image of the molten pool.A neural network model was established to estimate the penetration depth based on the welding current,pool width and seam gap.Also,the fuzzy logic technique was combined to promote the control accuracy of penetration depth.It was demonstrated that the proposed neural network could produce highly complex nonlinear multi-variable model of the GTAW process and thus it offered the accurate prediction of welding penetration depth.It was difficult to obtain the accurate models of the actuators,in that the torch drivers were extremely complex systems which had highly nonlinearity.In order to resolve this problem,a self-adjusting fuzzy controller to control the torch motion was proposed,which was used for seam tracking.The self-organizing artificial neural network algorithm was used to detect the weld position.The control parameters were adjusted on-line automatically according to the tracking errors so that the tracking errors could be decreased sharply.The experimental results showed that the proposed system yielded conspicuously controlling performance and provided an efficient approach to realize the intelligence of GTAW process.