High-performance phase-field lattice Boltzmann modeling for accurate prediction of melt pool dynamics in laser powder bed fusion
Please login to view abstract download link
In laser powder bed fusion, products are fabricated through repeated cycles of melting and solidification. The resulting solidification microstructure significantly influences the mechanical properties of the final part. Accurate prediction and control of this microstructure require high-fidelity modeling of melt pool dynamics. However, such simulations are computationally expensive, necessitating the use of efficient high-performance computing methods. In this study, we developed a high-fidelity phase-field lattice Boltzmann model along with an efficient high-performance computing framework to accurately capture melt pool dynamics during laser scanning. In this model, the solid-liquid [1] and gas-liquid [2] interfaces are described using the phase-field method, while melt pool flow is computed using the cumulant lattice Boltzmann method [3]. Laser scanning simulations using the proposed approach successfully reproduce key melt pool features, including keyhole formation, with high resolution. By utilizing multi-GPU parallel computing, large-scale simulations were completed within a practical time frame. This enabled detailed visualization of velocity and temperature fields in the melt pool, which have been challenging to capture using conventional methods.