Influence of statistical approaches on structural collapse potential assessment

Abstract

Structural collapse estimations could be influenced by alternative statistical methodologies that can be chosen throughout the seismic risk assessment of buildings, specifically in the generation of collapse fragility functions. This research evaluates the impact of selecting different statistical methods on the collapse potential metrics of a reinforced concrete dual wall-frame building as a case study. The explored methods are related to the identification and replacement of unusual observations in the simulated collapse data, and the selection of different probability functions (PFs) to characterize the collapse fragility. According to the results, this mid-rise building showed an acceptable Life Safety risk in accordance with ASCE 7 targets, regardless of the PF. However, the probability of collapse in 50 years obtained using the Weibull PF could be 2.6 times the value obtained using the lognormal PF. This observation is due to the larger probability of collapse values exhibited by the Weibull PF at small-to-medium Sa(T1) intensities that are associated to higher mean annual frequency of exceedance. In addition, it was also found that excluding extreme collapse values reduced the dispersion in the fragilities and it had a nonnegligible impact on the collapse probabilities for this case study building.