Multi-phase-field microstructure prediction for multiple layers and tracks in metal additive manufacturing

Microstructure prediction in laser powder bed fusion (LPBF) using the phase-field (PF) method has been limited to small regions with only a few layers due to high computational costs. In this study, we conducted PF simulations for microstructure prediction in an LPBF with 20 layers and four-track scanning employing two different scanning strategies. To achieve this, we utilized an efficient computational approach that integrates the moving frame algorithm with GPU-accelerated high-performance computing. The simulations revealed the three-dimensional grain morphologies formed through epitaxial growth across multiple layers alongside detailed two-dimensional cross-sectional microstructures. This method is anticipated to facilitate highly accurate microstructural predictions for various scanning strategies in future applications.