Quantifying time-dependent uncertainty in the BEAVRS benchmark using time series analysis

Abstract

Advances in computational capabilities have enabled the development of high-fidelity methods for large-scale modeling of nuclear reactors. However, such techniques require proper benchmarking and validation to ensure correct and consistent modeling of real problems. Thus, the BEAVRS benchmark was developed to legitimize the advancements of new 3-D full-core algorithms in the field of reactor physics. However, in order to address the issue of BEAVRS uncertainty quantification (UQ) of Uranium-235 fission reaction rate data, this thesis proposes a new method for measuring uncertainty that goes beyond merely conducting statistical analysis of multiple measurements at one given point in time. Instead, this work hinges on principles of time series analysis and develops a rigorous method for quantifying the uncertainty in using "tilt-corrected" data in an attempt to evaluate time-dependent uncertainty. Such efforts show consistent results across the four dierent methods and will ultimately assist in demonstrating that BEAVRS is a non-proprietary international benchmark for the validation of high-fidelity tools.