Evaluation of random-vibration procedures to estimate response spectral ordinates on soft soil sites from fourier amplitude spectra

Abstract

This study evaluates Random-Vibration Procedures (RVP) to estimate 5% damped response spectra of ground motions recorded on soft soil sites from Fourier Amplitude Spectra (FAS). The study makes use of 136 ground motions recorded in the soft soil Mexico City region during two recent large magnitude earthquakes. Seven different RVP are considered in this investigation consisting of four peak factor models, three strong ground motion duration models, and four non-stationary empirical correction factors. Spectral ordinates computed through conventional time integration were compared to approximate spectral ordinates computed using the FAS of the recorded acceleration time series in combination with RVP. A statistical error analysis was then conducted to determine how effective the various RVP were at estimating the exact pseudo-acceleration response spectra computed with response history analyses. It is found that most RVP commonly used today produce significant systematic overestimations of spectral ordinates at the fundamental mode of the soil deposit. Modifications to durations are proposed to be used in conjunction with various peak factors to reduce overestimations at the site period.