Optical coherence measurement-based penetration depth monitoring of stainless steel sheets in laser lap welding using long short-term memory network

Abstract

The industrial production site for laser lap welding of thin stainless steel plates puts strict requests on the level and fluctuation stability of penetration depth. Hence, the penetration depth monitoring is increasingly garnering attention. This research proposes an optical coherence measurement-based approach to monitor penetration depth utilizing machine learning in laser lap welding for stainless steel sheets. After the acquisition of the keyhole depth signal by the coherent light beam, it is found that there is a significant association relationship between the reconstructed keyhole depth obtained by empirical modal decomposition and the penetration depth curve, but meanwhile a penetration depth monitoring error also exists. Accordingly, based on the cross-correlation analysis and numerical simulation, it is revealed that the formation mechanism and sources of this error are related to the “bottom melt layer thickness”, “hysteresis property” and “multiple reflections”. On this basis, a long short-term memory network is built to memorize the historical information of the reconstructed keyhole depth for predicting the penetration depth at each moment. The experimental results demonstrate that the prediction model has high accuracy and good generalization ability, and thus effective monitoring for penetration depth can be achieved.