Multi-hazard system-level logit fragility functions

Fragility functions are used to represent the probability of failure of a structure or lifeline system conditional upon a hazard or set of hazards and are essential in the performance-based design process. Continuous lognormal damage fragilities are traditional, but recent formulations have implemented logit transformations from the family of generalized linear models for categorical data with a binary outcome (e.g., failure, no failure). In wind engineering, single hazard parameters derived from correlated variables (e.g., integrated kinetic energy, IKE) have been employed to indirectly include the effect of more than one hazard variable; however, even with more general hazard metrics, the lognormal formulation is still unduly restrictive for realistic fragility modeling. Using a statistical approach based on a logit formulation, a shift towards more robust fragility functions can be achieved. Because of its simplicity and ability to represent multiple predictor variables to improve the fitted model, this paper proposes use of the logit formulation of the fragility function at the system level for two or more simultaneous weather hazards. Successful applications of the model to characterize lifeline system-level fragility functions for electric power delivery during Hurricane Isaac in Louisiana and Hurricane Sandy in New York City using in-situ damage and hazard data are shown. While the results here are empirically derived, the modeling approach may be expanded for other structural systems subject to multiple loadings or demand variables.