Application of Bayesian optimal experimental design to reduce parameter uncertainty in the fracture boundary of a fuel cladding tube under LOCA conditions

成川 隆文; 山口 彰*; Jang, S.*; 天谷 政樹

Proceedings of 14th International Conference on Probabilistic Safety Assessment and Management (PSAM-14) (USB Flash Drive), 10 Pages, 2018/09

The reduction of epistemic uncertainty for safety-related events that rarely occur or require high experimental costs is a key concern for researchers worldwide. In this study, we develop a new framework to effectively reduce parameter uncertainty, which is one of the epistemic uncertainties, by using the Bayesian optimal experimental design. In the experimental design, we used a decision theory that minimizes the Bayes generalization loss. For this purpose, we used the functional variance, which is a component of widely applicable information criterion, as a decision criterion for selecting informative data points. Then, we conducted a case study to apply the proposed framework to reduce the parameter uncertainty in the fracture boundary of a non-irradiated, pre-hydrided Zircaloy-4 cladding tube specimen under loss-of-coolant accident (LOCA) conditions. The results of our case study proved that the proposed framework greatly reduced the Bayes generalization loss with minimal sample size compared with the case in which experimental data were randomly obtained. Thus, the proposed framework is useful for effectively reducing the parameter uncertainty of safety-related events that rarely occur or require high experimental costs.