A novel data-driven dimensional analysis framework for predicting melt pool morphology and porosity evolution in powder bed fusion.

As a cutting-edge additive manufacturing (AM) technique, powder bed fusion (PBF) has gained much attention in both academic research and practical applications. Revealing the mechanisms of PBF in a simple and physically interpreting way is imperative but also challenging, as PBF is a multi-scale process with comprehensive material properties and process parameters. Dimensional groups can shed light on the underlying physics involved in PBF, while traditional dimensional analysis approaches are limited in extracting unique dimensionless groups and measuring their importance. In this work, we integrated dimensional analysis with the active subspace method (ASM), a data-driven dimension reduction method, and developed a novel data-driven dimensional analysis (DDDA) framework for predicting the melt pool morphology and porosity evolution in PBF. Different from the conventional dimensional analysis, DDDA can explore more potential dimensionless groups while measuring their relative importance. Hence we identified those dimensionless numbers of most significant importance and calibrated the scaling laws for predicting melt pool morphology and porosity evolution. Furthermore, the correlation between track size and porosity formation, as well as how single-track keyhole can contribute to the bulk porosity, is also presented. With its performance demonstrated through good agreement with experimental results, it is believed that this proposed DDDA framework will be a useful tool for transforming our knowledge in the evolution of melt pool morphology, keyhole defects and porosity. [Display omitted] • A novel data-driven dimensional analysis (DDDA) framework for powder bed fusion (PBF) process is established. • Active subspace method combined with dimensional analysis is introduced into PBF process for the first time. • Dimensionless groups and scaling laws for melt pool morphology and porosity evolution are found by the DDDA framework. • Mechanisms behind the porosity evolution for both single/multi-tracks are identified. [ABSTRACT FROM AUTHOR]