A bayesian prognostic algorithm for assessing remaining useful life of nuclear power components

A central issue in life extension for the current fleet of light water nuclear power reactors is the early detection and monitoring of significant materials degradation. To meet this need nondestructive measurement methods that are suitable for on-line, continuous, in-plant monitoring over extended time periods (months to years) are needed. A related issue is then, based on a condition assessment or degradation trend, to have the ability to estimate the remaining useful life of components, structures and systems based on the available materials degradation information. Such measurement and modeling methods form the basis for a new range of advanced diagnostic and prognostic approaches. Prognostic methods that predict remaining life based on large crack growth, and phenomena that can be described by linear elastic fracture mechanics, have been reported by several researchers. The challenge of predicting remaining life for earlier phases of degradation is largely unsolved. This present paper will discuss the application of Bayesian prognostic algorithms applied to the early degradation- life prediction problem. A Monte-Carlo based algorithm is described for assessing the remaining useful life of materials from early stages of degradation. Results of applying the proposed algorithm to nondestructive measurements on fatigue specimens indicate the feasibility of remaining useful life estimation from early measurements of degradation.