DESIGN INSIGHTS RESULTING FROM THE COMPARISON OF U.S. NUCLEAR POWER PLANT COMPONENT BIRNBAUM IMPORTANCE MEASURES

摘要

The ability of a nuclear power plant to reach and maintain a safe shutdown condition is influenced by many factors, including those associated with a plant's operation, maintenance and design. Due to the confluence of these factors, coupled with variations in probabilistic modeling techniques, it is difficult to gain design insights as to the effectiveness of the designer's choice for a particular system or component configuration when comparing one power plant design to another. However, significant insights are obtainable if one can limit the operation, maintenance and modeling variations through the use of the U.S Nuclear Regulatory Commission (NRC) - sponsored Standardized Plant Analysis Risk (SPAR) models and focus on specific system functions through the use of component Birnbaum importance measures. These standard insights can then be validated through reviews of licensee probabilistic risk assessments (PRA). This paper describes the significant technical insights resulting from the comparison and grouping of components by attributes that have been shown to influence a component's risk importance. The attributes were determined through the investigation of the SPAR model cut sets in support of the implementation of the Mitigating Systems Performance Index. The investigation proceeded by selecting a component type and an associated failure mode (e.g., emergency diesel generator (EDG) fails to run), which typically equated to several basic events per plant (e.g., EDG A fails to run, EDG B fails to run, etc.) and investigating the resulting basic events for the fleet of U.S. plant designs. Each SPAR model was reviewed to determine the critical attributes influencing the component's importance. This was an iterative process. Often the number and type of attributes changed as the investigation progressed. Basic events having similar attributes were grouped and a resulting Birnbaum importance measure distribution was determined to validate the effectiveness of the grouping. This process was repeated for several component types. Once the initial grouping was completed, refinements were made based on limited reviews of licensee PRAs.