台形関数へのスペクトルフィッティングによる地震動加速度記録に基づく変位波形の推定

抄録

 A seismic ground displacement provides us the usable information to the analysis of the seismic source mechanism, the soil characteristics, the building damage, and so on. Especially, a huge residual displacement may cause a large response of long-period structures such as tall buildings and base isolated buildings.
 The accelerogram recorded by the servo sensor is often affected by the baseline irregularity due to the tilt motion of the sensor and the low-frequency noise, causing the difficulty of the estimation of the displacement including the residual component. In such case, the displacement is obtained by following two steps: the baseline change correction and the integration. In the first step, the irregularity of the acceleration baseline is corrected by eliminating the step function component in the frequency domain. Subsequently in the second step, the displacement is computed by integration using the real part of the Fourier spectrum.
 In the above mentioned scheme, however, usage of the low-cut filter may cause the deformation of the waveform and the underestimation of the residual displacement. In this study, the seismic ground displacement is obtained from the acceleration record by using a new technique. In the new method, the low-frequency noise is canceled replacing the original spectrum by that of the appropriate trapezoidal function, instead of using the low-cut filter. The parameters of trapezoidal function used are determined by the spectral fitting.
 Applying the method to the acceleration record of the 2016 Kumamoto Earthquake, the displacement is estimated accurately including the residual component. Particularly, the convergence of the displacement waveform in the coda part improves. In addition, the distribution of the residual displacement obtained by the method is almost consistent with that from GNSS observation.