AI-Driven Radiological Safety Assessments for Space Reactors in Hypothetical Reentry Scenarios

The development of space reactors signifies a transformative leap in modern exploration, providing sustainable energy solutions for extended missions and emergency scenarios. This study investigates the radiological impacts of a hypothetical space reactor accident during atmospheric reentry at an altitude of 70 km, a critical threshold for aerodynamic and thermal stresses that heightens risks of structural disintegration and radionuclide release. Employing the HotSpot code for radionuclide dispersion simulations, enhanced by a Long Short-Term Memory (LSTM) model, the research extends dose predictions to 321 km, exceeding HotSpot’s 200-km limit. The LSTM model demonstrated outstanding accuracy, achieving an R² of 99.5%, a mean absolute error of 4.9 × 10^-5, and a root-mean-squared error of 1.7 × 10^-4. It replicated HotSpot-calculated values while refining predictions for extended distances, estimating doses as low as 5.9 × 10^-5 kBq∙m⁻² at 321 km. The total effective dose equivalent (TEDE) at 0.03 km was 130 Sv on the first day and increased to 710 Sv after 55 years, reflecting the cumulative environmental radiation exposure, while organ-specific analysis showed that the liver absorbed 130 Sv on the first day, escalating to a cumulative 1300 Sv over the same period. Persistent contamination was observed up to 80 km, where TEDE values remained measurable at 0.036 Sv after 55 years. These results underline the dual necessity of mitigating both immediate and long-term radiological risks. By integrating advanced AI methodologies, the study provides dynamic, precise modeling to support robust safety protocols, environmental monitoring, and emergency preparedness strategies for space reactor operations.