Researchers from the Nagoya University in Japan, led by Yifan Sun and Hironobu Nakamura, have conducted a study that bridges experimental science and machine learning to advance the development of advanced nuclear fuels. Their work, titled “Thermophysical properties of spark plasma sintered UCo: a comparison with machine learning predictions,” was recently published in the Journal of Nuclear Materials.
The study focuses on the urgent need for accident-tolerant nuclear fuels that can enhance safety margins, particularly during loss-of-coolant accidents. Traditional uranium dioxide, while widely used, has a relatively low thermal conductivity, which is not ideal for maintaining fuel integrity under extreme conditions. To expedite the discovery of promising uranium compounds with higher thermal conductivity, researchers have turned to machine learning models. However, many potential compounds remain unexplored due to the challenges of handling radioactive materials.
In this research, the team fabricated uranium cobalt (UCo) and measured its high-temperature thermophysical properties. They found that the thermal conductivity of UCo, as predicted by machine learning, closely matched their experimental measurements. Although there were slight discrepancies, further analysis using SHAP (SHapley Additive exPlanations) values indicated that the model’s decision logic aligned with known physical trends.
The practical implications for the energy sector are significant. This study not only fills a gap in the reported thermophysical properties of UCo but also provides experimental validation for machine-learning-assisted screening of uranium compounds. This approach can accelerate the development of advanced nuclear fuels, enhancing safety and efficiency in nuclear power plants. By leveraging machine learning, researchers can more efficiently identify and test potential materials, reducing the time and resources required for traditional experimental methods.
The research was published in the Journal of Nuclear Materials, a reputable source for studies in the field of nuclear materials science. This work highlights the potential of combining experimental science with advanced computational techniques to drive innovation in the energy sector.
This article is based on research available at arXiv.