Human errors in ensuring the operability of standby systems

The study is a Finnish contribution to the project on "Human errors made in connection with maintenance" of the Nordic cooperation project on "Human reliability in complicated energy systems". In it the human factor problems in connection with the handling of standby components in nuclear power plants are discussed fron a more theoretical point if view. The study contains the following phases: (1) General psychological analysis of the test and maintenance activities, (2) Analysis of human error data in Swedish and Finnish nuclear power plants and (3) Case study at Loviisa. An attempt was made to define the particular nature of testing and maintenance activities. This would help on predicting the potential mechanisms of human errors in these activities. Three levels of analysis were separated: organizational, task, and performance control levels. In the case study typical problems were pointed out in each of them. On the basis of the qualitative and quantitative analysis of diesel generator testing human errors appear in the assessment of the status of the engine and in manipulating the engine. In particular the assessment proved to be critical. Informal organizational changes in the testing and repair units were recorded and interpreted as attempts to develop the knowledge bases for the assessment function. The positive effects of such spontaneous changes should be strenghtened through conscious measures aiming at a better integration of operation and repair. This would eventually make the acquisition of an adequate orientation basis more efficient. An enhancement in the assessment activity would promote a more productive and inspective testing style. This could be combined with partial testing rather than with mechanical regular testing. Concerning the errors in manipulating the engines (particularly during repair) the following hypothesis was formulated: Actions, that are secondary from the point of view of achieving the goal of the task, are lacking necessary feed back control and are thus prone to errors of omission. This hypothesis could, however, not be proved in this study because of lacking data of the goals and intentions of the opeators or repair men.