Review of machine learning applications in powder bed fusion technology for part production

Abstract

Additive manufacturing (AM) has become a viable option for production of industrial parts to meet the growing demands for customized components with complex geometry within a short lead time. Powder bed fusion (PBF) technology is often favored for its superior geometrical resolution and system stability. In the perspective of PBF machine users, the part production cycle can be broadly divided to three stages: the preparation stage, the printing stage and the postprocessing stage. The complexity in each stage gives rise to a challenging task for process control and quality assurance. The rising of machine learning in recent years sheds light on tackling this multi-factorial challenge in order to improve the overall performance of AM for part production. This article provides a review of machine learning techniques that are applied in or relevant to the part production cycle in PBF systems. The studies to date have showed segmented applications of machine learning techniques in different processes related to AM. To gain insight into the system behavior of PBF machines, more efforts could be put into constructing effective data representations and performing holistic analyses for the entire production cycle.