| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Microstructural Evolution and Material Properties Due to Manufacturing Processes: A Symposium in Honor of Anthony Rollett
|
| Presentation Title |
Enabling 3D Multiscale Materials Characterization Using Machine Learning |
| Author(s) |
Reeju Pokharel |
| On-Site Speaker (Planned) |
Reeju Pokharel |
| Abstract Scope |
We address the difficulty of characterizing three-dimensional (3D) mesoscale materials behavior across various spatial and temporal scales, emphasizing the need for faster data analysis capability to enable multiscale experiments at light sources. The current bottlenecks in the data inversion process make real-time feedback infeasible, thereby limiting our ability to pinpoint regions of interest during in situ experiments for guiding multiscale measurements. In this work, we demonstrate machine learning (ML) techniques to reconstruct crystal orientations from diffraction patterns, the refinement of ML algorithms for improved prediction accuracy, and the integration of physics knowledge for the robustness and generalizability of ML-based surrogate models. The overarching goal is to enable 3D multiscale material characterization studies at beamlines through immediate feedback by reducing the time required for data analysis and interpretation. Such an integrated approach is anticipated to improve our understanding of 3D mesoscale material dynamics. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Characterization |