| Abstract Scope |
This work presents AM-PROGRESS, a comprehensive simulation framework for large-scale sample modeling in powder bed fusion metal additive manufacturing. The framework couples thermal (FDM), microstructural (new PhaseLearn-CA), and mechanical (elastoplastic FEM) models to link print parameters to phase evolution and mechanical response. Alloy composition is integrated by adapting CALPHAD-informed thermophysical properties, transformation kinetics, and constitutive laws, enabling PhaseLearn-CA, a neural cellular automata framework, to rapidly predict phase fractions and grain evolution across spatiotemporally varying thermal histories. A Chimera mesh technique ensures seamless coupling between modules, and print-parameter-wise microscale thermomechanical simulations are used to generate multi-fidelity correction terms that enhance macroscale accuracy. A hybrid sampling strategy selects key points across voxelized STL models to build AI-ready datasets. This simulation setup can be calibrated using a few experimental samples and then scaled to generate a large, validated, physics-based database for predictive modeling, inverse design, and AI model training in advanced metal AM. |