PCM-based uncertainty reduction in support of model validation for depletion calculations. (1st December 2022)

• Record Type: Journal Article
• Authors: Huang, Dongli; Abdel-Khalik, Hany; Epiney, Aaron; Abdo, Mohammad; Mertyurek, Ugur
• Is Part Of: Annals of nuclear energy. Volume 178(2022)
• Publication Date: 2022-12-01
• Subjects: Model validation -- Uncertainty analysis -- Bias mapping -- Physics-guided validation; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2022.109332
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Nuclear model validation method, PCM, applying on isotopic depletion validation. Direct transferring of experiment uncertainty to application via mutual information. Non-intrusive forward-based algorithm to map biases and uncertainty. Non-informative prior allowed; applicable to nonlinear models. PWR to validate BWR lattice model with over 50% posterior uncertainty reduction. Abstract: This manuscript further develops a recent methodology, denoted by Physics-guided Coverage Mapping (PCM), to support model validation for neutronic depletion calculations. The overarching goal of model validation is to develop confidence in model predictions for the application of interest via fusion of both simulation results and measurements from scaled-down experiments, and whenever possible to improve predictions by explaining the observed discrepancies. This manuscript focuses on the isotopic depletion problem, that is how to improve the predictions of depleted fuel isotopic across the range of expected burnup based on a limited number of post-irradiation measurements. PCM employs an information theoretic approach, capable of directly transferring, i.e., without performing model inversion, biases and their uncertainties from the available measurements to the quantities of interest (QoIs), representing the isotopic concentrations at different burnup values and/or different irradiation spectra. It precludes the need for sensitivity coefficients and only requires forward model executions, and can be applied using non-informative priors, often required by Bayesian-based methods. This is achieved via a mapping kernel relating a number of predictor variables, the concentrations of single or multiple isotopes at certain burnup, to the QoIs, the isotopic concentrations at target burnup such as end of life. Proof-of-principle calculations are demonstrated using both representative PWR and BWR lattice models, where the goal is to employ measurements at given burnup value(s) from one lattice to predict the isotopic concentrations across burnup for the same or the other lattice. Results show 50% to 90% reduction in uncertainties of isotopic concentrations across burnup as compared to the prior uncertainty.