Thermal-hydraulic optimization of a closed-jet trans-critical hydrogen moderator with non-uniform heat sources using machine learning and multi-objective algorithms

Abstract

The optimization of heat transfer in transcritical hydrogen moderators, such as the Decoupled Poisoned Hydrogen Moderator (DPHM) at the China Spallation Neutron Source (CSNS), is critical for enhancing neutron flux and ensuring the safe operation of spallation neutron sources. This study addresses the unique thermal–hydraulic challenges posed by closed-jet impingement and non-uniform internal heat sources in transcritical hydrogen systems, which are essential for maintaining efficient neutron moderation and thermal stability. A three-dimensional computational model is developed to investigate the effects of key design parameters, including the flow baffle offset (L), inlet jet height (H), and hydrogen mass flow rate (m), on heat transfer and flow dynamics. Leveraging a genetic algorithm-optimized backpropagation (BP) neural network, this work introduces an innovative predictive framework to capture the complex nonlinear relationships between these parameters and system performance. Furthermore, a multi-objective optimization approach combining NSGA-III and TOPSIS is employed to identify optimal operating conditions that balance cooling efficiency and flow velocity constraints. The results demonstrate the critical role of jet impingement in mitigating thermal stratification and improving heat transfer, offering significant insights for the design and optimization of hydrogen moderators in advanced nuclear systems. This study not only advances the understanding of transcritical hydrogen behavior under non-uniform heat sources but also provides a practical, machine learning-enhanced methodology for optimizing thermal–hydraulic performance in spallation neutron sources.