Bayesian updating of fragility functions using hybrid simulation

The accuracy of fragility functions is critical for regional risk and loss estimations. This paper proposes a two-stage approach to generate improved fragility functions for engineering structures using field measurement and experimental data. In the first stage, the linear and nonlinear parameters of the bridge model are calibrated using measured earthquake responses and cyclic testing data; analytical fragilities are then generated with the calibrated model. In the second stage, a Bayesian updating approach is used to further update the derived fragilities using hybrid (analytical-experimental) simulation results. To illustrate the effectiveness of the proposed approach, fragility functions are generated for the Meloland Road Overcrossing Bridge considering four cases that represent an increasing level of data availability. A comparison of the four sets of fragility functions shows that appropriate calibration of bridge model is critical to the accuracy of the fragilities. In addition, hybrid simulation provides an economic and efficient way of validating and improving the accuracy of fragility functions through Bayesian updating.