Human errors related to maintenance and modifications

Abstract

The focus in human reliability analysis (HRA) relating to nuclear power plants has traditionally been
\non human performance in disturbance conditions. On the other hand, some studies and incidents have
\nshown that also maintenance errors, which have taken place earlier in plant history, may have an
\nimpact on the severity of a disturbance, e.g. if they disable safety related equipment. Especially
\ncommon cause and other dependent failures of safety systems may significantly contribute to the core
\ndamage risk.
\nThe first aim of the study was to identify and give examples of multiple human errors which have
\npenetrated the various error detection and inspection processes of plant safety barriers. Another
\nobjective was to generate numerical safety indicators to describe and forecast the effectiveness of
\nmaintenance. A more general objective was to identify needs for further development of maintenance
\nquality and planning
\nIn the first phase of this operational experience feedback analysis, human errors recognisable in
\nconnection with maintenance were looked for by reviewing about 4400 failure and repair reports and
\nsome special reports which cover two nuclear power plant units on the same site during 1992–94. A
\nspecial effort was made to study dependent human errors since they are generally the most serious
\nones.
\nAn in-depth root cause analysis was made for 14 dependent errors by interviewing plant maintenance
\nforemen and by thoroughly analysing the errors. A more simple treatment was given to maintenancerelated
\nsingle errors. The results were shown as a distribution of errors among operating states i.a. as
\nregards the following matters: in what operational state the errors were committed and detected; in
\nwhat operational and working condition the errors were detected, and what component and error type
\nthey were related to. These results were presented separately for single and dependent maintenancerelated
\nerrors. As regards dependent errors, observations were also made about weaknesses in audits
\nmade by the operating organisation and in tests relating to plant operation.
\nThe number of plant-specific maintenance records used as input material was high and the findings
\nwere discussed thoroughly with the plant maintenance personnel. The results indicated that instrumentation
\nis more prone to human error than the rest of maintenance. Most errors stem from refuelling
\noutage periods and about a half of them were identified during the same outage they were
\ncommitted. Plant modifications are a significant source of common cause failures. The number of
\ndependent errors could be reduced by improved co-ordination and auditing, post-installation checking,
\ntraining and start-up testing programmes.