One approach to including human errors in Probabilistic Risk Assessment (PRA) has sometimes been to simply assume a priori that they account for an arbitrary proportion of the failure risk (e.g., 80percent) and to further assess the effects of these errors on the probabilities and the consequences of the various failure modes. In this paper, it is shown that their estimated contribution to the risk of an accident can depend both on the depth of the analysis and on the system itself. Therefore the "80percent rule" is misleading and can yield wrong priorities among safety maasures. Instead, a systematic analysis can be undertaken to include these factors in a PRA for a specific system. The model permits assessing a wide spectrum of risk reduction options and avoiding double-counting. It is described and illustrated here by the case of the grounding of ships due to loss of propulsin.
展开▼