Mechanism analysis and feature extraction of arc sound channel for pulse GTAW welding dynamic process

Abstract

Arc sound signal has been proved to be an effective information for on-line monitoring and control for the quality of GTAW welding. Lots of studies focused on the sound source analysis. In this paper, an analysis of the generation and mechanism of sound channel during GTAW welding is completed, a method of establishing the equivalent model based on cepstral coefficient is proposed. The sound channel is expressed by complex cepstral coefficient through Z transform. The cepstral coefficient and spectral envelope of three different penetration states have been identified. Results show that the models of sound channel for full penetration, partial penetration and excessive penetration are clearly distinguishing the different fusion state of welding process. The sound channel for normal weld is more stable and balanceable than the abnormal weld. Thus, the cepstral coefficient for sound channel is proved to be an effective feature for the identification of dynamic welding process.