Bayesian Uncertainty Evaluation of Charpy Ductile-to-Brittle Transition Temperature for Reactor Pressure Vessel Steels

Abstract The irradiation embrittlement of reactor pressure vessel steels can be predicted using the ductile-to-brittle transition temperature (DBTT) shift obtained from Charpy impact tests. For the structural integrity assessment considering irradiation embrittlement, it is necessary to set margins for various uncertainties. It is important to understand what and how much factors contribute to the uncertainty. In the present study, a 34% credible interval value of Charpy DBTT at a 41J energy level (T41J) was evaluated by estimating the probability distributions of Charpy test data using Bayesian statistics. To fit the Charpy transition curves, a hyperbolic tangent with coefficients whose uncertainties depend on the test temperature was used. The probability distribution of T41J was estimated using Monte Carlo sampling and Bayesian inference. It was clarified that 34% of the credible-interval values of T41J before and after irradiation unchanged for base and weld metals when the number of specimens and test temperature were equivalent under un-irradiated and irradiated conditions. When the Charpy transition curve was determined by 12 specimens loaded in a surveillance test capsule, the estimated uncertainty of T41J was about 5 °C.