Prediction of penetration based on plasma plume and spectrum characteristics in laser welding

Abstract

The plasma characteristics generated in laser welding are closely related to the weld penetration state and the resultant weld properties. In this paper, a real-time monitoring platform was built to monitor the plasma plume and spectral in laser deep penetration welding of the H340LA high-strength steel tailor rolled blanks. The image processing method and the spectral processing method were used to investigate the correlation of the penetration on the plasma plume and spectral. The weld penetration state was divided into three categories: excessive penetration, moderate penetration and partial penetration. It was found that the penetration was mainly affected by centroid height, plasma area and intensity. The weld gradually transitioned from partial penetration, moderate penetration to excessive penetration with the centroid height, plasma area and spectral intensity decreased. When the centroid height was 2.42–4.19 mm, the area was 6.96–17.58 mm2, and the spectral intensity was 447.98–1453.95 a.u., moderate penetration weld can be obtained. A plasma plume-spectrum based multi-features Back Propagation (BP) neural network model was purposed to predict the weld penetration state with an accuracy of 97%, which is greatly higher than that of the single-feature neural network model. This work provides technical guidance for realizing real-time control of weld penetration.