Deveolpment of an intelligent operation supporting system for nuclear power plant

Abstract

This paper describes an integrated expert system, IOSS (Intelligent Operation Supporting System), which has been developed to support operators`` decisionmaking during the transients of the nuclear power plants. IOSS consists of three subsystems : ADES (Anomaly Diagnostic Expert System), ERGS (Emergency Response Guide System), and APS(Alarm Processing System). ADES diagnoses the anomalies of the nuclear power plants. In ADES, the shallow and deep knowledge approaches are combined to couple the merits of both approaches. Four types of knowledge are used according to the steps of diagnosis procedure: structural, functional, behavioral and heuristic knowledge. The structural knowledge is represented by three fundamental primitives, the functional knowledge by five kinds of functions. The behavioral and the shallow knowledge is represented by using the rules. As the inference strategies of ADES, the rule-based deduction and abduction are used for shallow and deep knowledge-based reasonings, respectively. The abduction is based on the human problem solving behavior in diagnosis tasks which have been modeled in ADES. ERGS monitors the status of critical safety functions, and generates possible success paths by using the structural knowledge of nuclear power plant. When the operators indicate a required safety system and its faulted components. The success paths are given to the operators by the order of unavailabilities of them. The required operators`` action to achieve them are, also, given to the operators. Finally, APS filters and suppresses unnecessary and potentially misleading alarms using the model-based reasoning and cause-consequence check rules. The alarms of nuclear power plants are classified into generalized and special alarms. For each type of alarm, general and special processing steps are applied. The priorities of alarms are adjusted according to the identified plant state. For the validation of IOSS, several validation studies have been performed u...