2D多尺度非线性地震速度成像

将遗传算法和单纯形算法相结合，得到了一种高效、健全的2D混合地震走时反演方法.把速度场划分为不同的空间尺度，定义网格节点上的速度作为待反演参数，采用双三次样条函数模型参数化，正问题采用有限差分走时计算方法，反问题采用多尺度混合反演方法.首先在较大的空间尺度内反演，然后减小空间尺度，将大尺度的反演结果作为次一级尺度反问题的初始模型，再进行混合反演，如此类推逐次逼近全局最优解.一个低速度异常体的数值模拟试验和抗走时扰动试验表明该方法是有效和健全的.我们将该方法应用到青藏高原东北缘阿尼玛卿缝合带东段上部地壳速度结构研究中，并与前人的成果进行了对比.

2D multiscale non-linear seismic velocity imaging

A hybrid algorithm of the global genetic algorithm (GA) and the local simplex method are proposed to form an efficient and robust 2D multi-scale non-linear seismic traveltime inversion method. The background velocity field is estimated at a spatially varying grid. The velocities at the nodes of the grid are regarded as inversion parameters. The model is parameterized by a bicubic spline function. The forward problem is calculated by use of the finite-difference meth od. The inverse problem is solved by the multiscale hybrid inverse method . First ,the inversion is done in large spatial scale, and then take its results as the initial value for inversion in next spatial scale. The inversion result will be closer to the global optimization. The method was tested to be efficient and ro bust by numerical simulating. An anomalous velocity area in the test model could be reconstructed and the result is stable when adding random noise to traveltimes. We also used the method to research the upper crustal structure features of A'nyêmaqên suture zone. At last, we compared our research results with that obtained by use of other method.