Blind Source Separation Based on Principal Component Analysis-Independent Component Analysis for Acoustic Signal During Laser Welding Process

Because of importance to safety and product quality, the online monitoring of the welding process performance has become a key issue for safety improvement. In order to guarantee the laser welding quality, acoustic monitoring based on microphone array was employed to the laser welding for sampling the acoustic signals during the whole welding process. However, in the hostile industry environments, the microphone array was limited by the multiple noise sources, including cooling-system and air-knife. In this paper in order to acquire more clearer acoustic Signal of the key-hole in the laser welding process blind source separation (BSS) based on principal Component analysis-Independent Component Analysis (PCA-ICA) is utilized. The non-Gaussian information of the key-hole acoustic signal can be extracted for defect detection and diagnosis. Meanwhile, spectrum analysis was applied to analyze the extracted signals including laser key-hole and cooling-system. By comparing the original cooling-system acoustic signal with the extracted cooling-system acoustic signal and analyzing the key-hole acoustic spectrum feature, the results showed that the extracted features were quite effective for laser welding detection monitoring. Meanwhile, the behavior of laser key-hole was explained by analyzing the acoustic signal extracted by the PCA-ICA algorithm. And the difference between welding blow-through defect and the normal welds of acoustic features can be distinguished through the BSS.