| About this Abstract |
| Meeting |
2022 TMS Annual Meeting & Exhibition
|
| Symposium
|
Mechanical Behavior and Degradation of Advanced Nuclear Fuel and Structural Materials
|
| Presentation Title |
NOW ON-DEMAND ONLY - Material Degradation Analysis through Machine Learning-based Information Extraction from Legacy SFR Metallic Fuel Performance Data |
| Author(s) |
Zhi-Gang Mei, Aaron Oaks, Kun Mo, Yinbin Miao, Logan Ward, Abdellatif Yacout |
| On-Site Speaker (Planned) |
Zhi-Gang Mei |
| Abstract Scope |
The FIPD database, the sodium-cooled fast reactors (SFR) metallic fuels irradiation & physics database, is established at Argonne National Laboratory and contains massive amounts of pin-by-pin post-irradiation examination (PIE) data with corresponding EBR-II operating condition parameters. Although the collected irradiation data have been widely accepted to qualify metallic fuels up to 10 at.% burnup, there are less extensive data at higher burnups. Machine learning (ML) approaches are capable of extracting complex relationships from the existing FIPD data to increase the confidence on the evaluation of fuel performance at high-burnups. As a case study for cladding degradation analysis, random-forest and deep neural network-based ML models were developed to predict the cladding strain with respect to a variety of irradiation parameters. Both models are found to provide more accurate predictions than empirical models. Finally, the potential application of ML to accelerate metallic fuel qualification using these legacy data will be discussed. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Nuclear Materials, Mechanical Properties |