UTILIZING DEGRADATION MONITORING FOR OPERATIONAL RISK ASSESSMENT

摘要

System/component degradations in nuclear power plants lead to reduction in system performance and plant economy, and further challenge safe operation of a plant by reducing the safety margins if they remain undetected. In many instances, it is hard to observe the signatures of degradation on the system behavior directly due to inefficient sensor placement, small disturbances as compared to measurement uncertainties, etc. Simultaneous multicomponent degradations may also mask the signatures of the degradations. For the cases when degradations in components/systems are detected and estimated, quantifying the operational risk associated with these degradations in that NPP in a timely manner is essential. We propose a degradation monitoring technique that is capable of detecting and estimating simultaneous multicomponent degradations for high dimensional and highly nonlinear systems. We present a degradation monitoring technique based on sequential Monte Carlo filtering with an adaptive Markov chain Monte Carlo (MCMC) step. This step works as a multiple hypotheses testing algorithm in which the hypotheses are constructed by utilizing a degradation database, which is compiled via past operational experience and manufacturer specifications. The adaptation scheme is based on a comparison of reproducibility of the limited number of measurements of the particles coming from the filter itself and from the degradation database to estimate the degradations in the components. A low-order model of a balance of plant of a boiling water reactor (BWR) is chosen as a demonstrative application. We show tests of our degradation monitoring algorithm for the estimation of nominal states, and multi-component degradations. In addition, we utilize the resistance-stress model taken from structural reliability analysis to evaluate the functional/performance failure probability of a degraded system and further assess its risk on plant operation.