時系列理論による強震動のシミュレーション

Abstract

This paper is concerned with procedures of synthesizing strong motion accelerograms, in whichthe earthquake motion is modeled as a nonstationary second order autoregressive moving average(AR-MA) process. Multi-dimensional AR-MA process is defined and the identification procedureis also developed for the case when the linear filter is excited by multi-dimensional non-stationarywhite noise. The filtering characteristics such as the natural frequency and damping factor ofmulti-dimensional system are detected by using recorded accelerograms. Using the simple timefunctions to model the filtering characteristics, the strong earthquake motions are simulated andcompared with the recorded one with the aid of power spectra.