About this Abstract |
| Meeting |
TMS Specialty Congress 2026
|
| Symposium
|
4th World Congress on High Entropy Alloys (HEA 2026)
|
| Presentation Title |
Learning Alloying Strategies from Small Datasets and Cheap Parameters to Increase Room Temperature Ductility of Tungsten-Based Refractory High Entropy Alloys |
| Author(s) |
Nick Beaver , Avik K. Mahata |
| On-Site Speaker (Planned) |
Nick Beaver |
| Abstract Scope |
Tungsten based refractory high entropy alloys show great potential for fusion and high temperature applications but suffer from limited ductility at room temperature due to a high ductile to brittle transition temperature. In this work, we applied machine learning to a small experimental data set to identify alloying strategies that increase room temperature ductility. A curated dataset of tungsten containing alloys was compiled from literature reports, and simple compositional and thermodynamic descriptors were derived to be model inputs. A random forest, support vector machine, and logistic regression model were trained to classify ductile behavior and interpret key factors controlling toughness. Model inspection revealed that valence electron concentration plays a dominant role in controlling ductility. The models were then applied to explore new compositional regions, revealing alloying trends that enhanced ductility. This approach demonstrates how interpretable machine learning can accelerate the design of more damage tolerant tungsten based refractory alloys. |
| Proceedings Inclusion? |
Undecided |