Real-time porosity monitoring during laser welding of aluminum alloys based on keyhole 3D morphology characteristics

Abstract

This research proposes a real-time porosity monitoring approach for aluminum (Al) alloys laser welding based on keyhole 3D morphology characteristics, realizing the prediction of the local porosity at each position of the weld. Building a multi-sensing signals monitoring platform consisting of a high-speed camera and a coherent optical system, the keyhole 3D morphology is measured. For the keyhole 3D morphological standard deviation (STD) features obtained by the algorithm of sliding window scan, the data processing based on ensemble empirical mode decomposition (EEMD) and principal component analysis (PCA) is adopted to acquire the input of an artificial neural network (ANN). The local porosity is acquired as the output of the ANN by scanning the weld with the sliding window and performing EEMD processing. A feedback mechanism is introduced into the ANN and the genetic algorithm (GA) is utilized to optimize the network parameters to improve the porosity prediction performance of the model. The prediction results show that the constructed optimized ANN can effectively and accurately monitor the porosity, and the generalization ability of the model is strong. In the corresponding relationship between the porosity and each STD feature, it is found that the formation position of the pore lags behind the fluctuation position of the keyhole, which can be explained by molten pool flow.