A diagnostic expert system for the nuclear power plant based on the hybrid knowledge approach

A diagnostic expert system called HYPOSS (hybrid knowledge-based plant operation support system), which has been developed to support operator decision-making during nuclear power plant transients, is described. HYPOSS combines shallow and deep knowledge to take advantage of the merits of both approaches. Four types of knowledge are used for the various steps of the diagnosis procedure: structural, functional, behavioral, and heuristic. The structural and functional knowledge are represented by three fundamental primitives and five types of functions, respectively. The behavioral knowledge is represented using constraints. The inference procedure is based on human problem-solving behavior modeled in HYPOSS. Event-based operational guidelines are provided to the operator according to the diagnostic results. If the exact anomalies cannot be identified and some of the critical safety functions are challenged, function-based operational guidelines are provided to the operator. For the validation of HYPOSS, several tests have been performed using the data produced by a plant simulator. The results showed the applicability of HYPOSS to the anomaly diagnosis for a nuclear power plant