Multisensor Fusion System for Monitoring High-Power Disk Laser Welding Using Support Vector Machine

In order to optimize the productivity of industrial manufacturing, a multisensor fusion system based on support vector machine (SVM) was researched to monitor and identify weld defects during high-power disk laser welding. Three different sensing technologies were integrated: 1) photodiode sensing for the monitoring of visible light radiation, which was generated from laser focus position; 2) ultraviolet and visible (UVV) sensing for plume and molten pool; and 3) visual sensing based on auxiliary illumination for the monitoring of the dynamic behavior of molten pool and keyhole. Time and frequency domains of the features that were extracted from the sensors constituted the eigenvector used for SVM classification. Experimental results showed that the integration of photodiode and visual sensing provided a more accurate and comprehensive estimation on the laser welding process. The proposed SVM-based approach has been proven to be efficient for inspecting defects in the laser welding process.