About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Fundamental Science of Microstructural Evolution and Phase Transformations: An MPMD/FMD/SMD Symposium in Honor of Peter Voorhees
|
| Presentation Title |
Tracking and predicting grain evolution with graphs |
| Author(s) |
Ashwin Shahani |
| On-Site Speaker (Planned) |
Ashwin Shahani |
| Abstract Scope |
Tracking the evolution of grains in time-resolved 2D/3D datasets is essential for understanding microstructural dynamics in polycrystalline materials, but remains challenging due to sample deformation, grain translation, and registration errors. We present a robust framework for grain tracking. Our approach reformulates the microstructure as a graph, encoding grains as nodes and grain boundaries as edges and uses graph matching to preserve topological consistency during tracking. Beyond accurate tracking, this graph-based representation can enable predictive modeling of grain growth outcomes. We benchmark our methods against traditional tracking approaches (volume-overlap and centroid matching) using both simulated (phase-field) and experimental (x-ray diffraction contrast tomography) datasets, including one with over 10,000 grains. In all cases, our graph based approach consistently outperforms other techniques, particularly under substantial structural evolution. This topology-driven approach not only enhances tracking accuracy but also has the potential to enable data driven insights into the mechanisms governing grain growth. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, Thin Films and Interfaces |