Research on welding penetration state recognition based on BP-Adaboost model for pulse GTAW welding dynamic process

Abstract

This paper proposed a new recognition model of analyzing the relationship between arc sound and penetration state. The experiment system is based on Robotic GTAW welding system with acoustic sensor and signal conditioner on it. The arc sound signal was firstly preprocessed to remove the influence of DC component and environmental noise. Then the features of arc sound signal were extracted and analysed in time and frequency domain. Finally, a new type of prediction model BP-Adaboost was built up to recognize different penetration state and welding quality through arc sound signal. The results showed that the new model had better prediction effect for the welding penetration state monitoring.