Development of an algorithm to discriminate between valid and false alarms in a loose-parts monitoring system

Abstract

Loose-parts monitoring system (LPMS) monitors loosened or detached parts and foreign parts inside the pressure boundary of a reactor coolant system. If any object is detected, the LPMS identifies the object's characteristics, and can contribute to improving plant safety, since it can identify loosened metal objects, which have a potential to cause severe damage to internal components of the steam generator chamber. The most significant problem of a traditional LPMS is the high false alarm rate. The most developed systems used more sophisticated methods for event identification. With these sophisticated systems, the false alarm rate could be reduced to below 1%. Even though the LPMS resulted in a false alarm rate of less than 1%, there is still a high false alarm rate when the unit increases or decreases power. As the unit increases power, the coolant starts to heat the metal structure, which causes false alarms. Plant operators should continuously identify all alarms, including false alarms, until the metal structure reaches thermal balance. It is difficult to discriminate between valid and false alarms, since the signal pattern by thermal shocks and structure friction is similar to that by loose metal impacts. The false alarm rate can be reduced to almost 0% by applying an algorithm to discriminate between the false and valid alarms. In this paper, an efficient algorithm is proposed to discriminate against signatures which are not related to loose-parts events, especially when the unit increases or decreases power. The algorithm can discriminate the signal pattern by the impact of loose parts against the signal pattern induced by thermal shocks and structure friction. The algorithm was validated in field tests by the proven automatic gun, and the false alarm rate was minimized.