Developing fragility surfaces for more accurate seismic vulnerability assessment of masonry buildings
Résumé
Usual methods for the vulnerability assessment of buildings often result in fragility curves, with respect to one intensity-measure parameter like PGA or spectral displacement. However, some recent studies have shown that other ground motion parameters may have a stronger influence on the building response. Besides, the seismic loading is very complex and describing it with one parameter may result in neglecting other significant characteristics of the ground motion (Seyedi et al., 2010). Thus, this study proposes a methodology to develop fragility surfaces, which express the probability of damage with respect to more than one parameter, in order to better account for uncertainties related to the hazard description. The proposed methodology is applied to a model of a two-story unreinforced masonry building (real structure tested on a shake table by Magenes et al., 1995) and the following steps are carried out: - modeling of the structure with TREMURI code (Lagomarsino et al., 2006) and performing static pushover analysis to check the accuracy of the model with the real structure; - development of a set of models with an aleatory distribution of mechanical properties; - based on Latin hypercube sampling, hundreds of dynamic analyses (using real and synthetic accelerograms) are performed and the damage level from each analysis is assessed; - a set of sensitivity studies on the ground motion parameters: clustering methods (segregation into groups of parameters) and correlation analyses help to select the appropriate couple of variable to use in the fragility surfaces; - application of the kernel estimation of density in order to get a non-parametrical distribution of damage probability and interpolation into fragility surfaces; The proposed framework permits us to choose a couple of ground motion parameters that are well correlated to dynamic response of the studied building and, in the same time, are less inter-correlated. The obtained uncorrelated intensity-measures are then used to build unbiased two-variables fragility functions and the couple of parameters (PGD and an acceleration level related to Arias intensity, for the studied building). A comparison between the one-parameter fragility curve and "slices" of the fragility surface (evolution of probability with respect to one parameter, while the other is kept constant) shows that the use of a second ground-motion parameter delivers a clearer definition of the vulnerability: the different "slices" can be seen as confidence intervals that can be of great help to public planners, especially in the case of very low probabilities for instance (i.e., extreme events).