A particle-based approach for the prediction of grain microstructures in solidification processes
Abstract
Grain microstructures are crucial to the mechanical properties, performance, and often lifetime of metallic components. Hence, the prediction of grain microstructures emerging from solidification processes at relevant macroscopic scale is essential to the design or optimization of new alloys and processing conditions. Yet, despite the broad range of multi-scale models proposed so far, all of them suffer from computational limitations, such that advances from computational and algorithm perspectives remain needed. Here, we present a novel approach for tracking crystallographic solidification grain envelopes capable of predicting competitive growth scenarios and columnar-to-equiaxed transitions for stationary grains. The model relies on classical assumptions and equations in use in several broadly used and thoroughly validated approaches (e.g. cellular automata). Yet, our approach defines the grain envelope using Lagrangian particles and tracks their evolution using an algorithm and an implementation relying on scalable libraries and using modern CPU/GPU architectures. The model is used to simulate several benchmarks of increasing complexity, and the results are compared to analytical, experimental, and numerical results from literature for the purpose of model validation. To highlight the applicability to real-world processes and the possibility of coupling the model with existing physics-based simulation tools, the model is also (one-way) coupled with a multiphysics laser-material-interaction model to simulate competitive grain growth during laser beam welding of steel.