Latency and mode of error detection in a process industry

Licensee event reports (LERs) from an industry provide important information feedback about safety to the industry itself, the regulators and to the public. LERs from four nuclear power reactors were analyzed to find out about detection times, mode of detection and qualitative differences in reports from different reactors. The reliability of the codings was satisfactory and measured as the covariance between the ratings from two independent judges. The results showed differences in detection time across the reactors. On the average about 10% of the errors remained undetected for 100 weeks or more, but the great majority of errors were detected soon after their first appearance in the plant. On the average 40% of the errors were detected in regular tests and 40% through alarms. Operators found about 16% of the errors through noticing something abnormal in the plant. The remaining errors were detected in other ways. There were qualitative differences between the LERs from the different reactors reflecting the different conditions in the plants. The number of reports differed by a magnitude 1:2 between the different plants. However, a greater number of LERs can indicate both higher safety standards (e.g. a greater willingness to report all possible events to be able to learn from them) and lower safety standards (e.g. reporting as few events as possible to make a good impression). It was pointed out that LERs are indispensable in order to maintain safety of an industry and that the differences between plants found in the analyses of this study indicate how error reports can be used to initiate further investigations for improved safety