About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Materials and Chemistry for Molten Salt Systems
|
| Presentation Title |
Advancing Molten Salt Thermal Property Models via ML-Predicted Structural Features of the Pair Distribution Function |
| Author(s) |
Jacob Numbers, Anthony Birri, Jackson Roy, Xingang Zhao |
| On-Site Speaker (Planned) |
Jacob Numbers |
| Abstract Scope |
Thermophysical properties of molten salts are deeply connected to the short-range order of molten salts as described by the pair distribution function (PDF), though PDF collection requires tuned interatomic potentials that limit the use of composition-generic property models reliant on the PDF. To advance the access and usability of PDF data for halide molten salts, existing PDFs in literature are compiled into an open-access database. The dataset of PDFs is used to train a machine learning model intended for generalized predictability across a wide range of salt structures. The model is used to predict key features of the PDF for use in thermophysical property models such as the structural coherence model of thermal conductivity. This work advances the practical applicability of physical models of microscale energy transport in molten salts, with the potential to provide increased resolution for heat transfer modeling in molten salt reactor applications. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Nuclear Materials, |