基于高斯函数的完全非平稳地震动多峰模型及其应用

The non-stationary feature of seismic ground motions has an important effect on the nonlinear dynamic response of structures, which makes it of great significance to accurately describe the non-stationary feature of seismic ground motions when generating artificial seismic waves. Based on the formula of spectral representation, the non-stationary feature of seismic ground motions was described by evolutionary power spectral density, which modulated the synthesized seismic waves by Gaussian function at each frequency section. General harmonic wavelet based Spanos-Tratskas method was adopted to calculate the evolutionary power spectral density of seismic ground motions. Combining peak pre-recognition technology and genetic algorithm, the fitting of evolutionary power spectral density(EPSD) model and target actual EPSD was achieved. A total of nearly 2 500 seismic ground motion records are selected from KiK-net earthquake databank to verify the applicability of proposed model, and several examples were provided. The results show that the proposed model can well fit the temporal envelope of EPSD at each frequency, especially at medium and high frequency components. The modeling of EPSD can extract the information of non-stationarity into identified model parameters, then the underlying pattern can be revealed by statistical analysis of model parameters. The proposed model provides a feasible way to study the non-stationary features of seismic ground motions.