Optimizing thermal stress distribution in heat-pipe-cooled microreactors using multi-physics data-driven methods

Heat-pipe-cooled microreactors represent cutting-edge nuclear energy systems, but they are faced with significant design challenges. The neutronic–thermal–mechanical interactions within these reactors are complex, and the excessive thermal stress in the core monolith threatens the structural integrity. We present an advanced multi-physics coupling method combining high-fidelity neutronic, thermal-hydraulic, and mechanical simulations. The model employs Monte Carlo neutron transport, finite element solid thermodynamics, and a heat pipe wall temperature model, coupled using Picard iteration. Leveraging this multi-physics coupling method, we have developed data-driven models utilizing artificial neural networks, which reduced each calculation time from 6 h to 4 min without sacrificing accuracy. To reduce the peak stress in the monolith, the non-dominated sorting genetic algorithm (NSGA-II) is applied to optimize key design parameters, such as fuel enrichment, active core height, and fuel rod pitch. An innovative heat-pipe-type optimization strategy was formulated, facilitating effective redistribution of core temperature and substantially reducing thermal stress without undermining key safety parameters or reactor performance. The optimization results showcased a 59% reduction in peak stress, with the maximum value decreasing to 68 MPa, thereby ensuring the stress levels remain within safe limits. This work suggests a viable solution for managing the thermal stress issues in heat-pipe-cooled microreactors.