Design Ground Motion for Lifeline System Taking into Account Fault Rupture Process

利用倒谱分析将地震波分解为系统响应函数和脉冲序列。基于误差最小化的几个准则，我们提出了一种利用识别的脉冲序列识别震源机制的方法，并对模拟的波形进行了分析，以检验所提出的反褶积方法的有效性。结果表明，脉冲到达时间的估计误差在两个采样间隔内，脉冲强度与脉冲的准确到达时间和估计到达时间之差成正比，利用El Centro阵列14个台站记录的加速度图识别1979年加州帝国谷地震的断层机制。由于解卷的14列脉冲包含有关故障机理的信息，因此，通过在破裂过程中选择合适的故障参数，使解卷后的脉冲列与理论计算值之间的误差最小，可以估计由破裂速度、破裂起点、破裂传播方向和上升时间等故障参数定义的破裂过程。在断裂面上识别出5个较大的小地震，证实了多次破裂过程，识别出的地震矩在断裂面上的分布与不同方法研究的结果吻合较好。这种解析算法导致了系统的非线性逆和辨识参数的不唯一性。为了克服这些问题，我们应用了贝叶斯规则，该规则提供了一种考虑先验信息影响的量化误差度量来识别参数。

The cepstrum analysis was applied to deconvolute a seismic wave into a system response function and a train of impulses. Based on several criteria to minimize error, we propose a method to identify the source mechanism using identified trains of impulses.Simulated waves were analyzed in order to examine the validity of the proposed deconvolution method. It was found that arrival time of an impulse was estimated within the error of two sampling intervals and intensity of impulse was proportional to the difference between the exact and estimated arrival time of the impulse.The acceleograms recorded at fourteen stations in El Centro Array are used to identify the fault mechanism of the 1979 Imperial Valley, California earthquake. Because deconvoluted fourteen trains of impulses contain information about the fault mechanism, the rupture process defined by such fault parameters as the rupture velocity, the starting point of rupture, the direction of the rupture propagation and the rise time can be estimated by minimizing the error between deconvoluted trains of impulses and theoretically calculated ones by selecting proper values of fault parameters in the rupture process. The multiple rupture process is confirmed by the fact that five major small events are identified on the fault surface.The identified distribution of seismic moment on the fault surface compares well with that investigated from different approach. The analytical algorithm led to the non-linear system inversion and non-uniqueness of identified parameters. To overcome these problems, we applied Bayes' rule which provided a quantitative error measure taking into account the effect of prior information for identifying parameters.