Robust fragility assessment using Bayesian parameter estimation

Analytical fragility curves corresponding to a prescribed limit state can be evaluated by propagating the various sources of uncertainty in the performance assessment of structures. This work proposes an efficient Bayesian parameter estimation method suitable for fragility curves based on a given probability model; in particular, the bi - parameter Log Normal probability distribution. This approach consists of Bayesian param eter estimation for the mean and the standard deviation of the L og Normal Fragility Model, given a limited (in the order of 10 to 50) number of structural analysis results for a specific structural performance variable denoted as Y. This leads to calculati on of the robust fragility and its percentiles or plus/minus standard deviation confidence interval underlying the number of simulations/realizations. The application of the robust fragility method is investigated in two cases: ( a ) vulnerability assessment to flooding for a class of non - engineered structures ; ( b ) seismic vulnerability assessment for an existing structure. . In case ( a ), the critical flooding height for collapse limit state is adopted as structural performance variable Y. In this case , the r esulting fragility curves are validated with respect to the results of Standard Monte Carlo simulation. In case ( b ), the spectral acceleration value corresponding to a critical demand to capacity ratio of unity is adopted as the structural performance para meter Y for the near collapse limit state