About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Advances in Multi-Principal Element Alloys V: Mechanical Behavior
|
| Presentation Title |
Combinatorial microstructure & stoichiometry-based and hybrid machine learning & domain knowledge for hardness mapping in a high-entropy alloy |
| Author(s) |
Mao-Yuan Luo, Ming-Yi Chen, Tu-Ngoc Lam, Wen-Jay Lee, E-Wen Huang |
| On-Site Speaker (Planned) |
Mao-Yuan Luo |
| Abstract Scope |
We developed a machine learning (ML) model using an artificial neural network to predict localized hardness in a Cu₁₅Ni₃₅Ti₂₅Hf₁₂.₅Zr₁₂.₅ high-entropy shape memory alloy. The model uses both alloy composition and microstructure as input. Elemental distributions—Ni-Hf-rich dendrites and Cu-Ti-Zr-rich interdendritic regions—were mapped using X-ray fluorescence. Local crystal orientations were obtained from X-ray nanodiffraction. Hardness measurements from nanoindentation were used to validate the predictions. We found that combining stoichiometry and microstructural features significantly improved the accuracy of predicted hardness maps compared to using composition alone. Adding more microstructure-related inputs can further enhance prediction performance. This approach clearly reveals the relationship between local microstructure and mechanical properties at the micrometer scale. Our method can be used to predict hardness variations caused by processing, offering a useful tool for alloy design and optimization. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
High-Entropy Alloys, Machine Learning, Mechanical Properties |