Researchers have developed a deep-learning model to predict subcooled flow boiling in isotope production facilities, where heat generated by particle beams hitting a target is removed by water channels. The model overcomes the challenge of real-world measurements due to high radiation levels during target irradiation. By analyzing temperatures and high-speed video of boiling in a mock apparatus, researchers validated the model, which can predict boiling in complex systems. This work, published in the International Journal of Heat and Mass Transfer, is crucial for isotope production facilities used in medical imaging and cancer therapy.

The model, adapted from a deep-learning tool originally designed for tracking biological cell activity, analyzes bubbles in water to indicate boiling, tracking bubble formation, size, and movement. By extracting critical bubble parameters from high-speed video, researchers developed a framework to predict a complete boiling curve and confirmed that current operations are below the predicted critical heat flux. This model can be expanded for full-scale modeling of complex multi-target and cooling channel geometries, benefiting other particle accelerator target applications and isotope production facilities worldwide.

The study, conducted at the Los Alamos National Laboratory Isotope Production Facility, highlights the importance of cooling systems in maintaining the proper function of target systems during irradiation. By leveraging deep-learning tools and experimental data, researchers have made significant progress in understanding and predicting subcooled flow boiling, ultimately improving the efficiency and safety of isotope production facilities.

Source link

Source link: https://phys.org/news/2024-06-deep-real-world-isotope-production.html