Particle beams hitting a target in an isotope production facility generate heat that is removed by water channels, leading to subcooled flow boiling. To test the cooling system’s limits, researchers built a mock apparatus due to high radiation levels during target irradiation. They used deep-learning tools to analyze data and validate a model predicting boiling in complex systems.
Isotope production facilities create isotopes for medical imaging and cancer therapy and require efficient cooling systems. Researchers at Los Alamos Neutron Science Center (LANSCE) used a deep-learning tool to track bubble behavior in water cooling channels to develop a model predicting cooling in target systems. This model allows for higher beam intensities without risking target failure, benefiting particle accelerator applications globally.
The research at LANSCE Isotope Production Facility aimed to determine maximum operating conditions of their target system, utilizing water channels to remove heat during irradiation. By analyzing temperature data and high-speed video of mock cooling channels, researchers were able to develop a model predicting a complete boiling curve using a deep-learning algorithm initially designed for biological cell detection.
Funded by the Department of Energy Isotope Program, the research at LANSCE Isotope Production Facility focused on improving cooling systems for isotope production facilities. By adapting deep-learning tools to analyze water cooling in the system, researchers created a model predicting boiling behavior, enabling higher beam intensities without risking target failure. This model can potentially benefit other particle accelerator target applications globally.