Failure detection and isolation methodology based on the sequential analysis and extended Kalman filter technique

A nuclear power plant operation relies on accurate and precise response of the monitoring system in order to assure a safety operational standard during the most predictable operational transients. The signal from the sensor are in general contaminated with noise and also with the randomic fluctuations making a precise plant assessment uncertain, thus with the possibility of erroneous operator decision or even with the false alarm actuation. In practice the noisy environment could even overcome the sensor malfunction misreading the plant operational status. In the present work a new failure detection and isolation (FDI) algorithm has been developed based on the sequential analysis and extended Kalman filter residue monitoring. The present methodology has been applied to both highly redundant monitoring systems and to non redundant systems where high signal reliability is required. (C.M.)