Seismic reliability assessment of infrastructure systems based on fragility models

Fragility models are widely used to describe the performance of components of infrastructure systems subject to earthquakes. Because the models cannot account for all variables involved in the physical process and because their parameters are estimated from a limited set of observations, they are subject to uncertainty. This uncertainty introduces statistical dependence among the model predications for individual components, which has been neglected in past applications. As demonstrated in this paper, this can lead to significant over-estimation of the reliability of redundant infrastructure systems. Here, a hierarchical regression model that accounts for statistical dependence among observations is applied to represent seismic fragility of equipment items in electrical substations. The resulting fragility estimates are compared to those obtained with conventional models. For redundant systems, differences of one to two orders of magnitude in the resulting probabilities of failure are observed.