Machine-Learning Based Modelling for AM Processes

Abstract

In this paper, additive manufacturing (AM) processes are categorized first, including its pre-and post-processes. Several high potential machine learning (ML) algorithms are then summarized briefly for modeling of AM processes. In particular, an integrative deep learning framework for multiscale coupled fields is discussed in detail for prediction of correlations among the AM process parameters, microstructures and mechanical properties of alloy and metal parts printed by AM, through integration of experimental raw data including the microstructure analysis and macro-scale mechanical testing of printed metal parts, handcrafted features with domain knowledge, and physics-based modelling input/output for laser powder bed fusion 3Dprinting process. Finally, potential ML applications are reviewed for modelling and simulation of AM process and relevant problems.