A Re-evaluation of Design Spectra for Seismic Damage Control

Keywords: Peak Ground Accelerations, Nonstationary Processes, Auto-Regressive Moving Average Models, Accelerographs, Stochastic Models, Damage Prediction, and Seismic Response Analysis.

Abstract: Seismic risk analysis for structural engineering purposes is primarily based on peak ground accelerations taken from earthquake records. Such an approach has a number of logical flaws and ignores a great deal of the information in measured data. This study investigates the use of simple, nonstationary ARMA models to represent underlying earthquake events. Measured earthquake records are considered to be random samples. Models are developed and samples of acceleration records are generated for four major events. Maximum displacement ductility demand, normalized hysteretic energy demand and a simple damage index spectra for bilinear and stiffness softening SDOF systems are computed for these samples of accelerograms. The sensitivity of demand spectra to ARMA model characteristics are also examined. It is concluded that for the events studied, simple ARMA models may be considered to capture most of the information contained in earthquake acceleration records insofar as nonlinear response spectra are concerned. It was found that, for each event, the average of the logarithms of displacement ductility and hysteretic energy demand for bilinear systems are very nearly linearly related to the logarithm of system period for SDOF systems.