Material degradation analysis and maintenance decisions based on material condition monitoring during in-service inspections

The degradation of the material in critical components is shown to be an effective measure which can be used to compute the risk adjusted economic penalty associated with different maintenance decisions. The approach of estimating the probability, with confidence interval, of the time that a prescribed degradation level is exceeded is shown to be practical, as demonstrated in the analysis of irradiated fuel cladding. The methodology for the estimation of the probability is predicated on the existence of a parsimonious and robust mixed-effects model of the evolution of the degradation. This model, in general, relates measured surrogates of the degradation level to computed or measured variables, which characterize the environment during the operating history of the component. We propose and demonstrate the efficacy of using an artificial neural network, constructed via a genetic supervisor, as an aid in developing the requisite mixed-effects model and testing its continued validity as new data are obtained.