Integrating Phase Field Modeling and Machine Learning to Develop Process-Microstructure Relationships in Laser Powder Bed Fusion of IN718

The current study aims to enhance our understanding of process-microstructure relationships in laser powder bed fusion (PBF-LB) of Inconel 718, utilizing computational fluid dynamics (CFD), phase field modeling, and machine learning. We developed a precise CFD model to simulate critical thermal dynamics in PBF-LB, addressing challenges posed by extreme temperature gradients and cooling rates. A high-throughput phase field model was then employed to predict microstructural evolution, focusing on the effects of solidification velocity on nucleation, grain growth, and component distribution. Additionally, we introduced the diffusion probabilistic field model (Diff-PFM), a machine learning model based on deep generative modeling. Trained on over 400 simulations, this model replicates intricate microstructural features and validates against experimental EBSD measurements. The integration of these sophisticated models establishes a robust framework for accurately predicting and controlling microstructure in LPBF processes, offering essential insights for the design of high-performance components and setting a new exemplar for additive manufacturing of metal alloys.