Diagnostic technique using a plant event data base

A diagnostic method with the functions of anomaly-cause identification and failure-effect prediction has been developed in order to support the operators of a nuclear power plant. The method consists of a plant event data base and a diagnostic section. The data base contains relationships represented by binary values computed from measurable process data under anomalous plant conditions. The diagnostic section consists of a status predictor which provides the predicted binary values, a supervisor which controls the diagnosis mode, an anomaly-cause identifier which identifies the cause of an anomaly and a failure-effect predictor which predicts the effects of component failure on the nuclear power plant. The method has been applied to the main subsystems of a PWR plant and the performance evaluated by use of a plant simulator. The results show that the method possesses high reliability, applicability to a real-time system, the ability to track plant operation and applicability to plant transients: these are the most important requirements for putting the method to practical use. (author)