Scalable Feature Tracking for Finite Element Meshes Demonstrated with a Novel Phase-Field Grain Subdivision Model

Abstract

The ability to identify features within finite element simulations and track them over time is necessary for understanding and quantifying complex behaviors as disparate as turbulent vortices in a flow field to microstructure evolution. We extend our previous research on feature identification in parallel unstructured meshes with the novel ability to maintain feature distinctness by dynamically remapping individual features to new simulation variables as the simulation evolves. We utilize this capability to drastically reduce the number of variables required in a simulation while maintaining the same fidelity as simulations without these reductions. We present this novel remapping algorithm and the corresponding implementation within the open-source Multiphysics Object Oriented Simulation Environment (MOOSE) framework. We demonstrate the utility of the method with a novel phase-field model of irradiation-driven grain subdivision in UO2. Grain population statistics are tracked over time, and a dynamically stable population of grains with a reduced size evolves. These results indicate that the small grain sizes observed in high-burnup UO2 can be explained by this mechanism.