About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Microstructure-Sensitive Modeling Across Length Scales: An MPMD/SMD Symposium in Honor of David L. McDowell
|
| Presentation Title |
Microstructure-Aware Bayesian Materials Discovery |
| Author(s) |
Raymundo Arroyave |
| On-Site Speaker (Planned) |
Raymundo Arroyave |
| Abstract Scope |
We propose a novel microstructure-sensitive Bayesian optimization (BO) framework to improve materials discovery by explicitly incorporating microstructural information. Unlike traditional approaches that primarily consider chemistry-process-property relationships, our method integrates microstructural descriptors as latent variables, establishing a robust process-structure-property mapping. Using active subspace methods for dimensionality reduction, we identify critical microstructural features, significantly reducing computational complexity while maintaining predictive accuracy. Enhanced Gaussian process modeling accelerates optimization, requiring fewer iterations and experiments to achieve optimal materials. We validate our framework through case studies, including the design of synthetic Mg₂SnₓSi₁₋ₓ thermoelectric materials as well as the deployment of the framework in real alloy discovery campaigns. This work suggests that since incorporating microstructure awareness improves the efficiency of Bayesian materials discovery, microstructure characterization stages should be integral to automated -- and eventually autonomous -- platforms for materials development. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, ICME, Machine Learning |