| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Composite Materials for Nuclear Applications III
|
| Presentation Title |
A Machine Learning Approach for Predicting Nuclear Fuel Performance With Solid Fission Products |
| Author(s) |
Denise Adorno Lopes, Rinkle Juneja, Matthew Kurley, Will Cureton, Christian Petrie, Andrew Nelson |
| On-Site Speaker (Planned) |
Denise Adorno Lopes |
| Abstract Scope |
A machine learning framework was developed to rapidly assess the effects of solid fission products on fuel performance and the phase stability of new fuel compositions with limited thermodynamic data. A comprehensive featurization strategy integrates 145 fission product attributes categorized into stoichiometry, elemental properties, electronic structure, and ionic properties to develop predictive models. Leveraging a dataset comprising approximately 152,000 compounds with materials properties computed via Density Functional Theory (DFT) from the Material Project database, we constructed a Random Forest model capable of predicting the formation energy of mixed compounds such as (U,X)O2 where X represents an array of fission products. The model is validated against DFT calculations for mixed uranium nitride compounds, (U,X)N, to evaluate the model's predictive accuracy. This framework supports the design of novel nuclear fuel composites for various reactor applications and could be extended to predict other relevant thermophysical properties using strategies such as cross-property transfer learning. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Nuclear Materials, Machine Learning, Modeling and Simulation |