About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Advanced Characterization Techniques for Quantifying and Modeling Deformation
|
| Presentation Title |
Learning Metal Microstructural Heterogeneity Through Spatial Mapping of Diffraction Latent Space Features |
| Author(s) |
M. Calvat, C. Bean, K. Vecchio, J.C. Stinville |
| On-Site Speaker (Planned) |
J.C. Stinville |
| Abstract Scope |
To leverage advancements in machine learning for metallic materials design, it is crucial to develop a data-reduced representation of metal microstructures that surpasses the limitations of current physics-based discrete microstructure descriptors. Capturing the spatial heterogeneity of microstructures at the different scales is necessary within such framework to accurately predict their properties. To address these challenges, we propose the physical spatial mapping of metal diffraction latent space features. This approach integrates EBSD point diffraction data encoding via variational autoencoders and the physical mapping of the encoded values. Together, these steps offer a novel means to comprehensively describe metal microstructures. We demonstrate this approach using wrought and additively manufactured alloys, showing that it effectively encodes microstructural information and, more importantly, enables the direct identification of microstructural heterogeneity that is not directly possible with physics-based models. Additionally, through ML-based inverse analysis, we identify the specific EBSD signals associated with targeted microstructural features. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, Additive Manufacturing |