Nonparametric Representation for Seismic Fragility Assessment of Earth Dams with Spatially Variable Soil Properties

Abstract

Seismic fragility is the conditional failure probability of reaching certain performance states under various intensities of seismic load, which is a significant part of the seismic risk and resilience assessment of structures and infrastructure systems. However, conventional seismic fragility curves are constructed based on the assumption of specific distribution models (i.e., a lognormal distribution), which might lead to a lack of fit and cannot reflect the true structures of data. Therefore, the validity of these assumptions remains questionable. In this study, a nonparametric approach that was based on dynamic reliability theory was employed to calculate the seismic failure probabilities under various intensities of seismic loads and seismic fragility curves were constructed for an earth dam, in which the spatial variability in the soil properties of the earth dam was properly considered by random field (RF) models. In addition, the results from the earth dams with homogeneous, random variables (RV), and RF models were compared to illustrate the effect of spatially variable soil properties on the seismic fragility of earth dams. The numerical results indicated that more comprehensive, reliable, and safe results could be acquired during the seismic fragility analysis of earth dams when the spatially variable dam properties were considered.