Monitoring of high-power disk laser welding of type 304 austenitic stainless steel based on keyhole dynamic characteristics

摘要

This paper presents an approach that uses multi-characteristic information fusion of a keyhole to obtain the welding status of high-power disk laser welding of type 304 austenitic stainless steel based on a back-propagation (BP) neural network. An infrared sensitive high-speed video camera was used to capture the dynamic image of a molten pool and keyhole. The centroid, area and perimeter of the keyhole were calculated using the image processing technique and were defined as the characteristic parameters of the keyhole, which were used as inputs to the neural network. The weldbeadwidth was consideredas aparameter reflecting the welding status, which was used as the output of the neural network. The effect of laser power on welding quality and status, weld depth and keyhole characteristics was also investigated. Experimental results show that the proposed method can effectively estimate the welding status when the laser power ranges from 2 kW to 10 kW.