COMPUTATIONAL MODELING AND PHYSICS-INFORMED MACHINE LEARNING OF METAL ADDITIVE MANUFACTURING: STATE-OF-THE-ART AND FUTURE PERSPECTIVE
Rahul Sharma, Y. B. Guo · Annual Reviews of Heat Transfer · 2022
Metal additive manufacturing (AM) processes, such as powder bed fusion and directed energy deposition, involves very complex thermal dynamics phenomena including repetitive rapid heating, fast solidification, and melt-back, which are not encountered in conventional manufacturing processes, such as casting, forging, and heat treatment. Understanding and predicting the thermal behaviors in metal AM remain one central challenge for printing high-quality metal parts. This paper presents a systematic analysis of the state of the art in physics-based computational modeling, data-driven machine learning (ML) modeling, and physics-informed ML modeling for metal AM processes. The physical laws, recent developments, and limitations in modeling metal AM processes are analyzed in depth for the physics-based computational methods including computational fluid dynamics, finite volume method, finite element method, lattice Boltzmann method, and discrete particle method. The complementary pure data-driven ML and physics-informed machine learning methods and applications in metal AM are also reviewed. The key findings, common challenges, research needs, and an outlook regarding new and emerging modeling methods are outlined, which may serve as a roadmap for the computational AM community for future synergy that enables smart metal AM.