About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
A Two-Step Physics-Informed Machine Learning Approach for NiTi-X Shape Memory Alloy Design |
| Author(s) |
Aysel Aysu Catal Isik, Shiyu He, Fei Xiao, Xuejun Jin, Sergio Gonzalez Sanchez, Enrique Galindo-Nava |
| On-Site Speaker (Planned) |
Aysel Aysu Catal Isik |
| Abstract Scope |
Shape memory alloys (SMAs) display superior thermoelastic properties due to the reversible martensitic transformation. Extensive alloy search using machine learning (ML) has been conducted based on the high-temperature NiTi-X SMA to improve shape recovery above 500°C. However, the predictive capability of ML models strongly depends on information quality, which restricts the existing design space to just a handful of elements tested in literature. This work proposes a two-step physics-informed ML framework to explore a larger and untested design space. Lattice parameters of main phases are estimated combining several ML models and geometric constructions. The B2-to-B19’ driving force is obtained from the deformation strain using a newly developed theory for martensite nucleation. The new physics-based parameters are combined with existing alloy datasets for new ML model adjustment. Predictions are compared against various experimental datasets, demonstrating that this new two-step ML approach is promising for discovering novel SMAs with enhanced properties. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, High-Temperature Materials, Computational Materials Science & Engineering |