Research On Uniform Hazard Spectrum And Conditional Mean Spectrum Based On Simulation Of Strong Earthquake Motions

The study on seismic hazard analysis and record selection in the buildings site without enough earthquake record data is very important to the seismic design of structures. As an important extension of probabilistic seismic hazard analysis (PSHA), the uniform hazard spectrum (UHS) is usually used as a target spectrum to select earthquake records. While the conditional mean spectrum (CMS) can provide the expected resonse sepectrum conditioned on occurrence of a target spectral accelation value at the period of interest, so it is another useful fool for selecting ground motions.In this thesis, the city Memphis TN in the USA is selected as a site location of interest to simulate the earthquakes that may occure in the future 1000 years. Based on the rate of occurrence, the distribution of magnitude, the location of focus, all the earthquakes that may occure near Memphis TN in 1000 years are simulated. The point source model and stochastic finite fault model are buillt, and the acceleration time history and pseudo-spectral acceleration spectrum are generated respectively. The point source model is AB95_BC model of the program SMSIM (3.1), while the stochastic finite fault model is a modified one with a normalized factor. All the records are analyzed with statistical methods using lognormal distribution in hypothesis testing. The curves of seismic hazard at T=0.1 & T=1.0s and UHSs with exceedance probability of 2% & 10% in 50 years are obtained, while the CMS corresponding to the UHS with exceedance probability of 2% & 10% in 50 years are also derived. Finally, the target spectrum is transformed from UHS to CMS, and 10 records which are compatible with CMS are selected from the database with a greed optimization technique.It is demonstrated in this thesis that the mean and standard deviation of the samples which are selected based on CMS are consistent with the target spectrum, and that the shapes of new samples which are dealt with the greed optimization technique are more consistent with the target spectrum.