About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Ceramics and Ceramic-Based Composites for Nuclear Applications III
|
| Presentation Title |
Design of High-Temperature Composite Radiation Shields Using Bayesian Optimization |
| Author(s) |
Byron Millet, Jim Steppan, Taylor Sparks, Tom Meaders, Lee Sorensen, Matt Coventry |
| On-Site Speaker (Planned) |
Byron Millet |
| Abstract Scope |
Radiation shielding (RS) that is stable at operating temperatures up to 650 °C is essential for future nuclear systems, including electromagnetic pumps for liquid sodium-cooled fast reactors and advanced microreactors. We demonstrate the application of Bayesian optimization to the RS design configuration and material selection for multi-layer composite radiation shields produced via filament winding with in situ potting. Composite radiation shields were fabricated using several radiation-shielding filler materials. An open-source, physics-based radiation simulation developed using GEANT4 for the evaluation of multi-layer RS configurations was coupled with Python-based machine learning algorithms to optimize composite RS materials and designs. Proof-of-concept composite shields, stable at 650 °C, showed improved neutron and gamma attenuation compared to an equivalent thickness of borated HDPE. The simulation was enhanced by incorporating CAD-based geometry to enable application-specific radiation shield optimization and streamline development by reducing the need for many expensive and complex experimental radiation exposure tests. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Modeling and Simulation |