A model for predicting vessel failure probabilities including the effects of service inspection and flaw sizing errors

Abstract

A numerical approach has been developed to predict the probability that a fabrication flaw in a reactor pressure vessel will extend by fatigue crack growth mechanisms and become a through-wall flaw. The fracture mechanics model treats the size of the flaw, the location of the flaw, and the parameters governing the fatigue crack growth rates as stochastic variables that are described by histograms that represent their statistical distributions. A latin hypercube approach forms the basis for efficient numerical calculations of vessel failure probabilities, in particular for those cases having very low probabilities that are not readily calculated by use of more conventional Monte Carlo simulations. A second aspect of the vessel failure model evaluates the benefits of in-service inspections at prescribed inspection time intervals and with prescribed nondestructive examination capabilities (probability of detection as a function of flaw size). A third aspect of the paper evaluates flaw sizing accuracy, and the impacts of flaw acceptance criteria. For representative values of flaw detection probability, flaw sizing errors, and flaw acceptance criteria, detection capability is the most limiting factor with regard to the ability of the in-service inspections to reduce leak probabilities. However, gross sizing errors or significant relaxations of current flaw acceptance standards could negate the benefits of outstanding probability of detection capabilities.