Bayesian Approach to Model Temperature Dependence of Charpy Absorbed Energy and Uncertainty Evaluation of Ductile-to-Brittle Transition Temperature for Reactor Pressure Vessel Steel

Abstract Embrittlement of reactor pressure vessel (RPV) steel caused by neutron irradiation has been evaluated using ductile-to-brittle transition temperature (DBTT) derived from surveillance tests (Charpy impact tests) during plant operation. For reliable structural integrity assessment of the RPV, incorporating adequate safety margins that take into account uncertainties inherent in surveillance Charpy impact tests are needed. In this study, a model to evaluate temperature dependence of Charpy absorbed energy variability using approximately 1900 datasets of unirradiated and irradiated materials manufactured in Japan and United States was developed. Next, probability distribution of Charpy ductile-to-brittle transition temperature at a 41 J energy level (T41J) was evaluated by estimating the probability distribution of Charpy test data using Monte Carlo sampling and Bayesian inference. From the detailed evaluation of the relationship between the number of specimens and T41J uncertainty, uncertainty of T41J was found to be almost the same in materials manufactured in Japan and U.S., and unchanged with neutron irradiation (no clear change in material inhomogeneity). Regarding product form on the other hand, uncertainty of T41J for base metal and weld metal was found to be comparable, indicating no significant difference, but the heat affected zone was shown to have large uncertainty.