Reference prior for Bayesian estimation of seismic fragility curves

One of the crucial quantities of probabilistic seismic risk assessment studies is the fragility curve, which represents the probability of failure of a mechanical structure conditional to a scalar measure derived from the seismic ground motion. Estimating such curves is a difficult task because for most structures of interest, few data are available, whether they come from complex numerical simulations or experimental campaigns. For this reason, a wide range of the methods of the literature rely on a parametric log-normal model. Bayesian approaches allow for efficient learning of the model parameters. However, for small data set sizes, the choice of the prior distribution has a non-negligible influence on the posterior distribution, and therefore on any resulting estimate. We propose a thorough study of this parametric Bayesian estimation problem when the data are binary (i.e. data indicate the state of the structure, failure or non-failure). Using the reference prior theory as a support, we suggest an objective approach for the prior choice to simulate a posteriori fragility curves. This approach leads to the Jeffreys prior and we prove that this prior depends only of the ground motion characteristics, making its calculation suitable for any equipment in an industrial installation subjected to the same seismic hazard. Our proposal is theoretically and numerically compared to those classically proposed in the literature by considering three different case studies. The results show the robustness and advantages of the Jeffreys prior in terms of regularization (no degenerate estimations) and stability (no outliers of the parameters) for fragility curves estimation.