Stochastic Simulation Of Seismic Ground Motions For Engineering Field And Applications

Simulation of ground motion technique can provide large quantities of data for seismic design in engineering field. Stochastic methods are widely used by researchers and engineers because they have clear concept, good simulation effect and are easy to use. Among these methods, point source based stochastic finite fault method is the most popular one. However, stochastic point source and finite fault methods are not faultless, for example there are uncertainties in parameter determination for the two methods and low computation efficiency for finite fault method. Moreover, Fourier based point source and finite fault have some limitations in simulating time-frequency nonstationarities and simulation effect is not good for ground motions that have strong nonstationarities. The present study investigated the similarities and differences between point source and finite fault methods, and focuses on the simulation of non-stationary ground motions based on wavelet packet method. The detailed works are as follows:(1)A stochastic point source model based on equivalent distance is proposed to consider non-uniform slip distribution of fault plane, according to differences between point source and finite fault method. The consistency between point source and finite fault method is analyzed then. The correctness and validity of the model is tested via simulation of acceleration for Wenchuan earthquake (2008) and simulation of isoseismal for Ludian earthquake (2014).(2) A simulation method for non-stationary ground motions is proposed based on wavelet packet decomposition and reconstruction technique. According to the characteristics of wavelet packet coefficients for real recordings, the model uses 6 parameters to separate time and frequency axis to investigate nonstationarities in time and frequency domain respectively. Simulation for real acceleration time history and comparison with existing stochastic model demonstrate the validity of the presented model.(3)Furthermore, a "Simulation-Fit" stochastic method for fitting design spectrum in seismic design field is proposed based on wavelet packet model. Taking advantage of high frequency resolution, wavelet packet coefficient can be adjusted according to design spectrum. Numerical example indicate that the method is flexible and has high fitting accuracy.(4) Regression model for wavelet packet model is proposed based on two-stage regression method and a ground motion database that contains 1222 horizontal recorded acceleration time history. The model links earthquake parameters with model parameters via inter-and intra-event residuals. Simulation for earthquake in the database and "prediction" for earthquake without the database indicates that the model has both deterministic and stochastic aspects. Comparison with existing ground motion prediction equations further demonstrates the validity of the regression model.(5) Validation of equivalent distance based point source method and wavelet packet based method is conducted from an engineering perspective. Ground motions for 4 earthquakes are simulated by EQSIM and WPSIM, and are compared with recorded acceleration time history through 4 structural response indexes, which demonstrates the effectiveness and practicability of the two method.