Decomposing seismic accelerograms with optimized window and its application for generating artificial fully non-Gaussian and nonstationary ground motion time histories

Abstract

In the present study, we propose a framework to simulate artificial seismic ground motion time histories for a given target record based on the time-frequency decomposition with an optimized window. The considered time-frequency decomposition technique is the S-transform (ST), where the optimized window is determined based on an energy concentration measure (ECM). For the evaluation of ECM for S-transform, we propose to use the time-frequency power spectral density function rather than the ST coefficients. Depending on the adopted assumption, the use of the iterative power and amplitude correction algorithm (IPAC) or the spectral represent method is suggested. The use of IPAC could also preserve the non-Gaussian characteristics of the target record. The proposed framework to simulate the artificial seismic ground motion time histories can capture the irregular pockets of power concentration in the time-frequency domain, and no assumptions on the functional forms of the spectra are required. The use of the proposed framework in simulating the time histories is illustrated by numerical examples, showing that the zero-upcrossing rate, the power distribution in time and in frequency, and the spectral acceleration of the simulated time histories match well their corresponding targets.