The optimal method for building damage fragility curve development

Abstract

A fragility curve is a primary component in the risk assessment, which is useful for evacuation planning, estimation of potential losses, and estimation of the damage to residential buildings caused by natural hazards. In general, a fragility curve represents the relationship between the probability of exceeding a specific damage state of a structure and natural hazard intensity. For determining such a curve, two parameters: the median and standard deviation are estimated. A fragility curve can be constructed using empirical data and analytical data. Numerical fitting data is used to develop the fragility curve. Various methods have been proposed using numerical fitting data to approximate the fragility curves. However, the most widely used methods for developing fragility curves are the least-squares method and the maximum likelihood method. In this present study, these two different numerical fitting data methods for fragility curve development are analyzed and compared. Basic assumptions and limitations of each method are also discussed. The building damage data used in all methods to derive the fragility curve is generated from hypothetical damage data assuming a lognormal distribution. Finally, the maximum likelihood method is proven to be optimal for developing fragility curve based on structural damage data.