Effects of long-period processing on structural collapse predictions

Abstract

We investigate the extent to which applying high-pass filters to ground motion records affects the collapse capacity of building models. We consider 26 ground motion records from seven large earthquakes and high-pass filter them with corner periods, T_c, ranging from 10 to 60 s. We perform incremental dynamic analysis on 9-, 20-, and 55-story steel moment-frame building models with fundamental periods of 1.88, 3.50, and 6.10 seconds, respectively. Even though filters with T_c ⩾ 20s have a minimal effect on the collapse capacities of the building models, we find that for a few motions, collapse capacities can increase by more than 50%, if T_c = 10 or 15 s, even for the 9-story models. We find that the collapse capacities with respect to raw, uncorrected records are generally similar to those of the tilt-corrected versions, indicating that removing long-period noise with high-pass filters can make collapse predictions less accurate, if T_c  < 20 s.