Real-time monitoring of abnormal conditions based on Fuzzy Kohonen clustering network in gas metal arc welding

Abstract

A real-time monitoring system based on through-the-arc sensing is developed for detecting abnormal conditions in gas metal arc welding. The transient signals of welding voltage and current during the welding process are sampled and processed by statistical analysis methods. It is found that three statistical parameters (the standard deviation, variance, and kurtosis of welding current) show obvious variations during the step disturbance, which is intentionally introduced into the T-joint test pieces by cutting a gap in the vertical plane. A Fuzzy Kohonen clustering network (FKCN) is put forward to monitor the abnormal conditions in real-time. Ten robotic welding experiments are conducted to verify the real-time monitoring system. It is found that the correct identification rate is above 90%.