Vision detection and feature extraction of molten pool behavior in powder laser cladding

pool monitoring is the basis of forming process optimization, playing a pivotal role in improving the formation quality. A machine vision monitoring system was set up for the laser cladding process. Based on the OpenCV software library, a method combining K-means image segmentation and dual threshold Otsu image segmentation was proposed, which realizes the accurate distinction between molten pool and plume with the accuracy of extracting geometric features of molten pool contour up to 95%. By designing an orthogonal experiment of laser cladding Ti-6Al-4V metal powder, typical cladding samples were selected to analyze the width and shape change of the weld pool in time domain, obtaining the fluctuation rule of the weld pool of the cladding layer. The results show that the fluctuation frequency and amplitude of the molten pool are affected by the process parameters and their combinations, and gradually stabilize with the progress of the cladding. Abnormal fluctuations in the molten pool during the cladding process are conducive to the positioning and identification of cladding defects, helping to optimize process routes.