Criteria for Assessing the Quality of Nuclear Probabilistic Risk Assessments

The final outcome of a nuclear Probabilistic Risk Assessment (PRA) is generally inaccurate and imprecise. This is primarily because not all risk contributors are addressed in the analysis, and there are state-of-knowledge uncertainties about input parameters and how models should be constructed. In this thesis, we formulate two measures, risk significance (RS) and risk change significance (RCS) to examine these drawbacks and assess the adequacy of PRA results used for risk-informed decision making. The significance of an event within a PRA is defined as the impact of its exclusion from the analysis on the final outcome of the PRA. When the baseline risk is the final outcome of interest, we define the significance of an event as risk significance, measured in terms of the resulting percentage change in the baseline risk. When there is a proposed change in plant design or activities and risk change is the final outcome of interest, we define the significance of an event as risk change significance, measured in terms of the resulting percentage change in risk change. These measures allow us to rank initiating events and basic events in terms of relative importance to the accuracy of the baseline risk and risk change. This thesis presents general approaches to computing the RS and RCS of any event within the PRA. Our significance measures are compared to traditional importance measures such as Fussell-Vesley (FV), Risk Achievement Worth (RAW), and Risk Reduction Worth (RRW) to gauge their effectiveness. We investigate the use of RS and RCS to identify events that are important to meet the decision maker's desired degree of accuracy of the baseline risk and risk change. We also examine the use of 95 th confidence level acceptance guideline for assessing the adequacy of the uncertainty treatment of a PRA. By comparing PRA results with the desired accuracy and precision level of risk and risk change, one can assess whether PRA results are adequate enough to support risk-informed decisions. Several examples are presented to illustrate the application and advantages of using RS and RCS measures in risk-informed decision making. We apply our framework to the analysis of the component cooling water (CCW) system in a pressurized