Seismic fragility evaluation of nuclear power plants based on bayesian inference

Yamaguchi, Akira

Seismic fragilities of equipment and systems are evaluated in a seismic probabilistic safety analysis (PSA). The seismic fragility is defined as the failure probability and its uncertainty at various ground acceleration levels. One evaluates the seismic fragility efficiently by making most use of available information in the seismic PSA. This study is related to a Bayesian inference method to reflect the information to the seismic fragility evaluation. Information that can be used in the Bayesian inference is (1)seismic experience data, (2)expert judgment on the seismic fragility, (3)seismic test data. The generic fragility database was used as the prior fragility that was to be updated. The acceleration level of the seismic experience and test data is much lower than the median value of the seismic capacity level of equipment. However, it is useful to cut down the lower tail of the fragility curves thus the modeling uncertainty. As the results, it is has been found that the annual frequency of failure that is calculated by the convolution of the fragility with seismic hazard curves is reduced considerably. Furthermore, the author has proposed the concept of entropy to quantify the value of the information used in the Bayesian inference. The combined use of the Bayesian inference and the entropy is a useful method to propose a cost-effective seismic test for the fragility evaluation. Findings obtained from the seismic PSA are taken into consideration in achieving more reasonable safety design. It should be noted that the findings are strongly affected by the uncertainty. Hence the reduction is necessary of the uncertainty of the contributive equipment from the viewpoint of seismic safety. Specific analysis and test of the contributive equipment is an effective way of reducing the uncertainty.