Parameterized Fragility Assessment of Bridges Subjected to Hurricane Events Using Metamodels and Multiple Environmental Parameters

Abstract

A vulnerability assessment of bridges subjected to hurricane events is beneficial to policy makers prioritizing their critical assets for inspection, maintenance, and retrofit. The proposed parameterized fragility assessment effectively builds a relationship between the hurricane hazard intensity/bridge parameters and bridge response by implementing a metamodeling approach, and produces a fragility surface which describes the probability of failure in terms of two hazard intensity measures: wind speed and water elevation. These environmental parameters are practical indicators of hazard intensity. The independent variables comprise bridge material/geometric parameters as well as hazard intensity parameters, and the output parameters represent a binary classification of bridge failure or no-failure states. The proposed approach is used in a vulnerability assessment of Georgia’s coastal bridges using three-dimensional finite-element models. One thousand five hundred statistical samples are generated using a range of parameters determined from common bridge types to study seven metamodels. The results indicate that the random forest model is the most suitable in predicting hurricane-induced failure, and that fragility surfaces are successfully created for two hazard intensity parameters.