Fragility curves production for steel structures by seismic improvement of the high-dimensional model representation method

Fragility curves are utilized to evaluate the probability of exceeding the damage index for structures ‎exposed to seismic hazards. The Monte Carlo simulation method, which involves generating ‎random numbers, is computationally expensive for calculating fragility curves. To address this ‎issue, several methods have been proposed to produce fragility curves at a reduced computational ‎cost. This study presents a method that enhances the seismic representation of high-dimensional ‎models to generate accurate fragility curves for steel structures while significantly decreasing ‎computational costs. This method selects uncertain variable values based on the results of initial ‎incremental dynamic analyses. The fragility curves are divided into three zones, and an equation is ‎proposed to estimate mean damage values associated with the boundaries of these zones. ‎Additionally, polynomial response functions were generated to estimate the fragility curves. The ‎proposed method is applied to generate the fragility curves for three steel structures, one with 4, 9, ‎and 12 stories. Fragility curves are generated for four damage levels: non-structural damage (DS1), ‎structural retrofitting required (DS2), intensive structural damage (DS3), and collapse (DS4). The ‎resulting fragility curves are compared with those generated by the Monte Carlo simulation method ‎and other existing methods. The comparison demonstrates that the proposed method achieves ‎fragility curves with a significant decrease in computational costs compared to the Monte Carlo ‎method, while also exhibiting higher accuracy than other methods. The maximum error of the ‎proposed method is approximately 20%, whereas Cornell's and the conventional HDMR methods ‎exhibit errors of up to 80% and 60%, respectively. The errors of other methods increase ‎significantly for fragility curves associated with high damage levels and 9 and 12 story steel ‎structures, where nonlinear structural behavior is pronounced. In contrast, the increase in error is ‎not significant in the proposed method. The findings of this study can be utilized to assess the ‎seismic impact of various stochastic factors, such as random eccentricity or loading-related ‎parameters, on the vulnerability of steel structures.‎