初始模型对地震走时层析成像的影响及解决方法.doc

初始模型对地震走时层析成像结果的影响分析及对策 张小磊，张东 (武汉大学 物理科学与技术学院，武汉430072 摘 要：常规地震走时层析算法对初始模型有较强的依赖性，初始模型与实际模型偏差较大时，需要经过很多次的迭代才能得到较好的反演结果，甚至反演结果不收敛。针对这一问题，本文分析了不同初始模型的反演结果存在较大差异的原因，提出了一种对常规层析算法进行改进的方法，以提高算法对不同初始模型的一致收敛性。研究发现，模型中没有射线覆盖或者射线覆盖较少的地方是影响算法对初始模型依赖性的关键，因此本文提出使用动态插值方法改进反演过程，根据迭代次数和结果收敛情况自动选择合适的插值方法。实验表明，改进的算法不仅提高了层析成像算法的对不同初始模型的一致收敛性，也提高了反演结果的精度。 关键词：初始模型；一致收敛性；动态插值 中图分类： 631.4 文献标志码： The analysis and the solution of the initial model problem to travel-time tomography Zhang Xiaolei, Zhang Dong (Wuhan University, School of Physics and Technology, Hubei Wuhan 430072, China) Abstract:Conventional seismic travel-time tomography algorithm is highly dependence on initial model, it always takes many times of iteration to get better inversion results, even only get a no convergence inversion results when the initial model and actual model make a great difference. In order to solve this problem, in this paper we analyzed the cause that why the inversion results with different initial models make a great difference, and proposed a method to improve the conventional tomography algorithm, which can improve the uniform convergence of different initial models. Studies have found that in the model where there is less rays or no ray coverage is the key to affect the algorithm dependence on initial model, this paper puts forward using dynamic interpolation method to improve the inversion process, select the appropriate interpolation method automatically according to the number of iterations and the convergence situation of results. Experiments show that the improved algorithm can not only improve uniform convergence of the tomography algorithm to different initial model, but also improve the accuracy of the inversion results. Keywords:initial model; uniform convergence; dynamic interpolation 0 引言 走时层析算法可以很好的反演出地质模型的速度，已被广泛的应用于地震勘探中[1,2]。在地质模型反演时，需要选取一个初始模型，若初始模型选取合适，收敛速度较快，同时反演结果与实际模型非常接近，若初始模型选取不当，则反演需要迭代很多次，甚至收敛到的反演结果与实际模型偏差很大。初始模型的选取关系到层析成像算法的效率甚至结果的正确性[3,4,5]。 针对初始模型对层析成像的影响，很多学者做了如何选