Nonstationary Stochastic Models of Earthquake Motions

Eight strong-motion accelerograms recorded on firm ground and at moderate epicentral distances are studied for the purpose of developing a reasonable stochastic representation of the time history of ground accelerations during strong earthquakes. The results indicate that over durations which are of interest in structural response calculations, nonstationary random processes are needed for the description of the records. A nonstationary Gaussian filtered shot-noise process is examined for this purpose. The member functions of the mathematical model and their associated linear response spectra are compared with the corresponding information obtained from the recorded accelerograms. On this basis, a nonstationary Gaussian filtered shot-noise process with a second-order filter is proposed for stochastic simulation of strong-motion earthquakes.