Multigrid Mode Verification for Reduced Order Modeling

New generations of advanced nuclear reactors will address several issues with the current technology. The development of reactor designs based on new concepts requires fast-running modeling and simulation tools that can capture multiphysical phenomena within reactors. These tools should be fast, accurate, and easy to operate. A new fast-running methodology is proposed for the physics analysis of pebble bed reactor designs, considered the most promising advanced nuclear reactor design. An established methodology will be used to track the location and trajectories of fuel pebbles.Based on the data generated bypebble tracking,a series of high-fidelity neutronics analyses will be carried out to train a neural network. Once trained, the neural network will be used to quickly obtain physical properties of the reactor and act as an accurate replacement for slow, detailed analysis. The design tool will then rely on this neural network to perform fast-running, middle-fidelity simulations. The feasibility of the project will be confirmed. The proposed neural network will be designed and trained on a canonical pebble bed reactor design. The accuracy of the method in capturing the physics of the nuclear reactor will be investigated. Finally, the fitness of the method to design applications will be studied. If successful, the tool developed under this project will be usable by over 50 US companies that are actively developing advanced reactor designs. The US Departments of Energy and Defense will also be able to use this tool to address national needs.