Use of Uncertainty Importance Measures to Complement Risk Importance Measures in PSA

摘要

Fussel-Vesely (FV) importance and Risk Achievement Worth (RAW) are two commonly used risk importance measures for ranking the importance of basic events in probabilistic safety assessment (PSA) of nuclear power plants. They are based on the point estimates of the probabilities of basic events. The uncertainties which are existed in basic events are not considered in FV and RAW. The identification of the contribution of the uncertainty in a basic event to the uncertainty in the overall model risk is done through uncertainty importance analysis, which can help us focus on the events more information of which is needed to reduce the variability of the overall model risk. By using two fault tree models as examples in this study, we compared the rankings of basic events based on FV and RAW with those based on two uncertainty importance measures. The results show that the basic events which contribute the most to the uncertainty in the overall model risk might not the most important risk contributors. Since uncertainty importance measures and risk importance measures, such as FV and RAW, provides different views of the importance of a basic event to the overall model risk, it is argued that they should be used together to prioritize information where further resources should be allocated.