The Role of Mechanistic Modeling and Multiscale Uncertainty Propagation for Accelerated Fuel Qualification

Fuel qualification for nuclear reactors is a notoriously diAicult burden for new reactor designers. The lack of available data, combined with the high cost of obtaining new integral experimental data, limits the agility of the nuclear industry when diverting from traditional oxide fuel forms. Fuel performance simulations have helped fill this gap with some success, starting first with supporting design calculations, and slowly making inroads in support of licensing and safety calculations. In an eAort to help strengthen the benefit of fuel modeling in data-poor materials, multiscale modeling can be used to help bridge the gap where adequate atomistic data is obtainable, and mechanistic models exist for which the lower-length scale data can plug in. For example, we will show that multi-scale simulations have provided a wealth of data for fission gas release and swelling predictions in an otherwise empirical dominated space, connecting atom-to-atom behavior to safety characteristics of the fuel such as plenum and contact pressures. More recently, data science tools have helped contextualize these results through the use of uncertainty quantification and calibration. We will show that when done correctly, multiscale uncertainty propagation can be the unifying tool between fuel performance models, experimental results, and fuel qualification by elevating simulations into the same language, i.e., uncertainty.