A Critical Review of Machine Learning Methods Used in Metal Powder Bed Fusion Process to Predict Part Properties

Metal Powder Bed Fusion (M-PBF) technique is one of the popular branches of Additive Manufacturing (AM). One of the biggest challenges in M-PBF is understanding relationship between processing parameters and produced part’s mechanical properties. In this review paper, recent M-PBF and Machine Learning (ML) studies are comparatively investigated to guide the scientific community in selecting right ML algorithm to predict and optimize the mechanical properties of the parts produced by M-PBF technique. In this context, theoretical background of M-PBF techniques are discussed in terms of processing parameters and mechanical properties. Constraints on M-PBF processes are examined and possible solutions are studied. ML theory is briefly reviewed and various ML algorithms are investigated regarding their applicability and validity for M-PBF processes. Popular Design of Experiments (DOE) methods are reported. Future trends and suggestions on M-PBF techniques are discussed.