Statistical analysis using the Bayesian nonparametric method for irradiation embrittlement of reactor pressure vessels

Takamizawa, Hisashi ; Ito, Hiroto; Nishiyama, Yutaka 

To understand neutron irradiation embrittlement in high fluence regions, statistical analysis using the Bayesian nonparametric (BNP) method was performed for the Japanese surveillance and material test reactor irradiation database. The BNP method is essentially expressed as an infinite summation of normal distributions, with input data being subdivided into clusters with identical statistical parameters (such as mean and standard deviation) for each cluster to estimate shifts in ductile-to-brittle transition temperature (DBTT). Clusters typically depend on chemical compositions, irradiation conditions, and the irradiation embrittlement. Specific variables contributing to the irradiation embrittlement include the content of Cu, Ni, P, Si, and Mn in the pressure vessel, neutron flux, neutron fluence, and irradiation temperatures. It was found through numerous examinations that the measured shifts of DBTT correlated well with calculated ones. Data associated with the same materials were subdivided into the same clusters even if neutron fluences were significantly disparate among the results. This indicates that slowly developing or late-onset embrittlement mechanisms were not evident in the present study.