A method of hybrid convolutional neural networks (CNNs) was proposed for powder-bed fusion (PBF) process monitoring in this work. The proposed method can learn both the spatial and temporal representative features from the raw images automatically based on the advantages of the CNN architecture. The results demonstrated the superior performance of the proposed method compared with the traditional methods with handcrafted features. The overall detection accuracy of four process conditions, e.g., overheating, normal, irregularity and balling, can be up to 0.997. In addition, it is found the temporal information for PBF process monitoring by the vision detection of the process zone (including melt pool, plume and spatters) is significant. As the proposed method can save image processing steps, it simplifies the procedure on feature extraction. This makes it more suitable for online monitoring application.

As a highly promising technology in additive manufacturing, the laser powder bed fusion has only limited application due to its low reproducibility. In this study, the image information of the 316L specimen after laser scanning and powder paving of each layer was acquired by a complementary metal–oxide–semiconductor industrial camera. The important features were selected, extracted and quantificated by analyzing the tensile test results. Finally, combined with the laser power, the quantified features were as input of a Gaussian process regression model based on optimization algorithm of grid search to predict the tensile strength of 316L specimen. The results show that the quantized image features have a significant improvement on the regression effect, and the coefficient of determination (R ²) is improved from 63% to 90.57% compared to using only the laser power as input.