Graph attention-based knowledge reasoning for mechanical performance prediction of L-PBF printing parts

Owing to the high-costing and complexity of traditional mechanical performance testing for laser-powder bed fusion (L-PBF) forming parts, it is imperative to propose a novel performance prediction method for printing parts based on data and knowledge reasoning. In this paper, a multi-layer graph attention-based knowledge reasoning is proposed to consider the influencing parameters of the L-PBF process to predict the mechanical properties of L-PBF printing parts. The proposed prediction approach can be considered an effective solution to address the challenges associated with the performance prediction of L-PBF printing parts. Furthermore, the proposed prediction approach has been demonstrated through a practical case study on the performance prediction of the L-PBF process. This case is thoroughly analyzed using graph attention and fully connected neural networks, incorporating both simulation and experimental data to validate the prediction results of L-PBF printing performance under complex process parameters. The common results of L-PBF part performance have been predicted to provide many detailed perspectives and future possible trends. Finally, a detailed conclusion has been given.