Human performance contribution to risk in nuclear power operating events

摘要

This paper presents results of a study of the contributions of human performance to risk in operating events at commercial nuclear power plants. The Nuclear Regulatory Commission (NRC) sponsored the study, which was performed by the Idaho National Engineering and Environmental Laboratory (INEEL). The NRC Accident Sequence Precursor (ASP) Program and the Human Performance Events Database (HPED) were used to identify safety significant events in which human performance was a major contributor to risk. Conditional core damage probabilities (CCDPs) were calculated for these events using Systems Analysis Programs for Hands-on Integrated Reliability Evaluation (SAPHIRE) software and Standardized Plant Analysis Risk (SPAR) models. Forty-eight events described in licensee event reports and augmented inspection team reports were reviewed. Human performance did not play a role in 11 of the events, so they were excluded from the sample. The remaining 37 events were qualitatively analyzed. Twenty-three of these 37 events were also analyzed using SPAR models and methods. SPAR analyses were not possible for the other fourteen events, because they involved operating modes or conditions outside the scope of the SPAR models. The results showed that human performance contributed significantly to the analyzed events. Two hundred and seventy human errors were identified in the events reviewed and multiple human errors were involved in every event. Latent errors (i.e., errors committed prior to the event whose effects are not discovered until an event occurs) were present four times more often than were active errors (i.e., those occurring during event response). The latent errors included failures to correct known problems and errors committed during design, maintenance, and operations activities. The results of this study indicate that multiple errors in events contribute to the probabilistic risk assessment (PRA) basic events present in SPAR models and that the underlying models of dependency in HRA may warrant further attention.