Multi-objective optimization of DU bed for hydriding performance and heat management based on data-driven surrogate model

The Storage and Delivery System (SDS) of ITER is responsible for the controlled charging and discharging of hydrogen isotopes, requiring reliable and efficient metal hydride storage technology. This study numerically investigates hydrogen-loading behavior in a depleted-uranium (DU) bed to identify optimal configurations for tritium storage. A multilayer-perceptron (MLP) surrogate model is developed to predict absorption performance, and multi-objective optimization using NSGA-II is conducted to simultaneously maximize hydrogen absorption efficiency and minimize bed volume. Results show that elevated operating temperatures raise equilibrium pressures and thereby suppress hydrogen absorption, while improved thermal regulation enhances absorption kinetics, as reflected in the temperature and H/M distributions. The resulting Pareto front provides optimal design trade-offs between performance and compactness. These findings highlight the central role of thermal management in metal-hydride-based hydrogen storage and offer practical guidance for designing high-performance tritium beds in fusion systems.