The application of neural networks for the control of industrial arc welding

The use of automatic closed loop control is well established in all areas of manufacturing industry. New methods for measuring system variables, data processing and process control are being sought to improve system efficiency. Skilled welders are able to subconsciously monitor a manual arc welding process by listening to the sound and repositioning the electrode in response to a change in arc noise. This paper describes the real time monitoring of acoustic emissions from an automated submerged arc welding process and the application of Neural Networks to predict the point of instability of the process variables.