Knowledge-based modeling of operator response for severe-accident analysis

Studies of severe accidents in light water reactors have shown that operator response can play a crucial role in the predicted outcomes of dominant accident scenarios. Although computer codes such as MAAP are available to predict the thermal-hydraulic response, substantial knowledge about plant practices and procedures is needed to make reasonable assumptions about operator response. Based on the thermal-hydraulic state of the plant, symptom-oriented procedures provide general guidance to the operators, who then take one of several possible actions. The paper pictures this process as a feedback loop that relies heavily on the judgment of the individual safety analyst. The ability to more explicitly model the procedural guidance and operator response can help close this analytical loop and improve the overall integration and consistency of severe accident analysis. An object-oriented model for operator response characteristics and symptom-oriented procedures was developed using the NEXPERT OBJECT expert system shell. This prototype system reads MAAP transient output files and determines the instructions and operator response characteristics that are implied by the observable plant variables. A limited set of boiling water reactor (BWR6) emergency operating procedures (EOPs) was formulated as a rule set, and pattern-matching techniques were used to generate message queues for display and reports