On the importance of uncertain factors in seismic fragility assessment

This paper addresses the definition of importance measures for helping the modeller to detect the factors on which to focus modelling activity and data collection in seismic fragility analysis. We study sensitivity measures consistent with the decision-support criteria of interest, namely, the (mean) fragility curve and the “High Confidence of Low Probability of Failure” (HCLPF) value. The importance measures are obtained analytically for the EPRI safety factor method, which is nowadays used worldwide for seismic risk assessment of nuclear plants. We illustrate and discuss the use of both variance-based and CDF-based importance measures in the application to two case studies, the first analytical and based on the EPRI method, the second numerical.

► Identifying the most and least relevant factors is essential to manage uncertainty in seismic fragility assessment.
► We introduce important measures that look at entire fragility curves and HCLPF as well as variance of the capacity.
► Analytical and numerical case studies illustrate the methodology.