MODELING CHARPY IMPACT ENERGY PROPERTY CHANGES USING A BAYESIAN METHOD

Surveillance schemes that monitor the effect of neutron irradiation on reactor pressure vessel materials employ Charpy impact specimens that are periodically withdrawn and tested as a function of test temperatu re. The resulting Charpy impact absorbed energy curves have been model ed by a three-parameter relationship with the same functional form as the Burr distribution function. The parameters of the Burr distributio n function have been represented as a function of irradiation variable s: dose and temperature. The method of maximum likelihood estimation h as been used to obtain estimates of model parameters together with the ir standard errors. A method is presented for the evaluation of uncert ainties in Charpy impact energy curves and temperature shifts resultin g from irradiation damage. A sampling approach, based on Bayesian infe rence and employing Markov chain Monte Carlo (MCMC) simulation, has be en used to quantify the uncertainties. The main idea of MCMC sampling is to generate the distributions of the model parameters by successive random sampling from the multivariate normal distribution. Illustrati ve results are presented for irradiated submerged are weld metal.