Approximate Bayesian Computation of the occurrence and size of defects in Advanced Gas-cooled nuclear Reactor boilers

A campaign of visual inspections was undertaken in 2014 on a specific component of thirty-two Advanced Gas-cooled nuclear Reactor boilers at Hartlepool and Heysham A power stations, as a consequence of an anomalous reading previously obtained in a routine ultrasonic test. While the presence of a crack in Heysham A boiler 1D1 was confirmed, no further defect was detected; not all components could be inspected in all of their eight sections (octants), however, because of the very unfavourable conditions in which the work had to be performed. The occurrence and size of defects in the uninspected octants are here inferred on the basis of models of crack initiation and growth fitted, by Approximate Bayesian Computation, to the results of the campaign. The probability of failure of one or more components in an extreme seismic event is also inferred. - Highlights: • The occurrence and size of defects in thirty-two AGR boiler spines are estimated. • The estimate is based on the outcome of a recent campaign of visual inspections. • Models of defect initiation, growth and detection are fitted to the data. • The model fitting is performed by Approximate Bayesian Computation. • The probability of component failure in an extreme seismic event is estimated.