Machine learning to optimize additive manufacturing parameters for laser powder bed fusion of Inconel 718

Abstract

Approximately 3600 samples have been printed to characterize the build parameters for laser powder bed fusion (L-PBF) fabrication of Inconel 718. The tested samples connect pore formation to part orientation, part location and the use of recycled powder. These data serve as the basis for development of a Random Forest Network machine learning (ML) model capable of two-way modeling of process–property and process–structure relationships. These results show how common procedural steps in the setup and execution of L-PBF effect porosity, particularly the formation of keyhole and lack of fusion (LOF) defects, and how data collection, processing, and validation can expose even subtle connections between input features and output parameters using a general ML framework.