b'Accelerating utilization ofIntegrating high-fidelity nuclear fuel performance modeling into fuel performance modelingmultiphysics simulation enhances reactor safety and economy.using artificial intelligence A s the heart of nuclear reactors, nuclear fuels undergo complicated thermo-mechanical-chemical degradation that pose constraints to reactor operations. Accurate and reliable fuel performance modeling is therefore essential to ensure the safe and economic operation of nuclear reactors. However, emphasis has been placed on transitioning from empirical correlations to physics-informed models in fuel performance analysis, and mechanistic fuel performance models are generally subject to deficiencies due to absent information PROJECT NUMBER:about model parameters or missing physics. In addition, fuel performance analysis 22P1066-005FP has traditionally remained a stand-alone process in engineering scale multiphysics TOTAL APPROVED AMOUNT:modeling even though it provides essential information on fuel deformation and $230,000 over 2 years failure risks that are directly associated with reactor safety. PRINCIPAL INVESTIGATOR:To enhance the reliability of fuel performance modeling, we developed efficient Yifeng Che Bayesian inference methods for inverse uncertainty quantification. More specifically, this project focused on Bayesian inference confined to constrained domain, which CO-INVESTIGATORS: remained a challenging problem for commonly adopted sampling algorithms. Ryan Stewart, INL Spherical measures were introduced into the Hamiltonian Monte Carlo framework to Som Dhulipala, INL implicitly handle the boundary conditions. The spherical measures were integrated Wen Jiang, INL with computationally efficient Hamiltonian Monte Carlo based algorithms such James Tompkins, Radiant as No-U-Turn Sampler and physics-informed machine learning algorithms like COLLABORATOR: Hamiltonian neural networks. Efficiency and accuracy of the developed algorithms Argonne National Laboratory significantly improved over conventional algorithms, as demonstrated on multiple synthetic analytical problems and tentatively applied onto the tristructural isotropic particle fuel.The second part of this project focused on integrating high-fidelity fuel performance modeling into the MOOSE framework to provide essential insights into the fission product transportation, thermo-mechanical responses, and failure risks of nuclear fuels during reactor operation. Such high-fidelity multiphysics simulation allows for more realistic evaluation of the safety margin, which is of significance to reactor operation and design optimization. The multiphysics simulation was developed for a high temperature gas cooled microreactor, covering the full-core neutronics calculation, thermal-hydraulics analysis in the coolant channel, and the fuel performance modeling of each single tristructural isotropic particle. This work served as the first-of-a-kind demonstration of integrating high-fidelity fuel performance modeling in BISON into the multiphysics coupling scheme under the MOOSE framework, opening the door to fine-grained analysis and optimization of nuclear reactors in the future. 54'