Statistical Analysis and Stochastic Simulation of Ground-Motion Data

The time variations of the root-mean-square accelerations, the autocorrelation functions, and the power spectral density functions of 12 strong-motion earthquake accelerograms are analyzed. The results indicate that: (i) strong-motion accelerograms of sufficiently long duration are stationary in the rms sense, (ii) the stationary rms acceleration is a good measurement of earthquake intensity, and (iii) the autocorrelation and power spectral density functions of strong-motion accelerograms resemble those of a narrowband process. Based upon these results, a method of determining the transfer characteristics of a site is introduced. A procedure for generating a filtered, gaussian stationary process to simulate ground motions is developed, and two applications of this simulation procedure illustrate its significance.