時系列理論による強地震動特性の推定

Abstract

This paper is concerned with a procedure for synthesizing strong motion accelerograms, inwhich the earthquake motion is modeled as the nonstationary second-order autoregressive movingaverage (AR-MA) process. This means that the synthesized accelerograms can be obtained asthe output of one degree of freedom system excited by nonstationary white noise. The filteringcharacteristics such as the natural frequency and damping ratio of the one d-o-f system can becalculated from AR-MA parameters determined by the two-stage least-square method from thesequence of data segmenting the earthquake record. In the first least-square stage, the autoregres-sive parameters are estimated by using the concept of maximum entropy method proposed byBurg.Using the record obtained at 1968 Tokachi-Oki and 1971 San Fernand earthquake, the timedepending filtering characteristics of the one d-o-f system are examined in detail and classifiedinto typical four or five groops.