Feature extraction of molten pool for laser welding quality real-time inspection

This paper proposed an image feature extraction method for laser welding molten pool inspection based on cellular neural network. TC4 titanium alloy thin plates were welded by Nd:YAG pulsed laser. A coaxial machine vision system was designed to acquire molten pool images. An auxiliary lighting source was employed to improve the molten pool image quality. By analyzing molten pool images, the welding defects such as fenestration or insufficient depth were identified. These results can be used as a feedback signal for laser power control. Experimental results showed that the proposed method can be used to improve laser welding quality.