CHARPY IMPACT ENERGY DATA - A MARKOV-CHAIN MONTE-CARLO ANALYSIS

To assess radiation damage in steel for reactor pressure vessels in th e nuclear industry, specimens are subjected to the Charpy test, which measures how much energy a specimen can absorb at a given test tempera ture before cracking. The resulting Charpy impact energy data are well represented by a three-parameter Burr curve as a function of test tem perature, in which the parameters of the Burr curve are themselves dep endent on irradiation dose. The resulting non-linear model function, c ombined with heteroscedastic random errors, gives rise to complicated likelihood surfaces that make conventional statistical techniques diff icult to implement. To compute estimates of parameters of practical in terest, Markov chain Monte Carlo sampling-based techniques are impleme nted. The approach is applied to 40 data sets from specimens subjected to no irradiation or one or two doses of irradiation. The influence o f irradiation dose on the amount of energy absorbed is investigated.