MELTING POOL SIMULATION FOR METAL ADDITIVE MANUFACTURING

Abstract

Additive manufacturing of metals, particularly LPBF of Inconel-718 superalloy, involves complex thermal and fluid dynamics that determine build quality. Motivation: Optimizing LPBF remains challenging due to the stochastic nature of powder beds and the difficulty of experimentally observing melt pool behavior. Methods: This work uses a multi-physics phase field method implemented in COMSOL Multiphysics to simulate melt pool formation in Inconel-718 during LPBF. Realistic powder bed configurations are generated to represent actual random packing of powder (including a distribution of particle sizes), and an idealized uniform powder layer is also modeled for comparison. The phase field approach allows explicit tracking of the evolving metal–air interface as the laser heat source melts the powder. Key Results: Simulations reveal the transient melt pool geometry and internal fluid flow under a moving laser heat source. The melt pool in a realistic, polydisperse powder bed reaches approximately ∼40μm in depth, ∼100-120μm in width, and ∼200μm in length. The melt pool in a idealized, monodisperse powder bed reaches approximately ∼50μm in depth, ∼120-150μm in width, and ∼400μm in length. It was found that the idealized powder bed overestimates melt pool dimensions compared to the realistic powder bed. Molten metal circulates within the melt pool at velocities ranging from 0.3 to 0.5 m/s. These flows redistribute heat and molten material, leading to a more uniform thermal profile in the idealized case and a somewhat irregular pool shape in the realistic case. Contribution: The phase field model successfully reproduces key phenomena of LPBF melt pool dynamics - including interface evolution, melt pool morphology, and convective flow - using material properties of Inconel-718. This study provides new insight into how powder bed randomness influences melt pool characteristics, thereby improving the fidelity of melt pool simulation for metal additive manufacturing and offering a tool to better predict process outcomes.