基于迭代再加权小二乘的地震资料稀疏反演方法硕士.doc

分类号 密级 UDC 编号 中国科学院研究生院 硕士学位论文 基于迭代再加权最小二乘的地震资料稀疏反演方法 申请学位级别 硕士 学科专业名称 固体地球物理学  Seismic Data Sparse Inversion Method Based on Iterative Reweighted Technique By Yuan Yi-ming A Dissertation Submitted to Graduate University of Chinese Academy of Sciences In partial fulfillment of the requirement For the degree of Master of Science Lanzhou Center for Oil and Gas Resources，Institute of Geology and Geophysics，Chinese Academy of Science 签字日期：年 月日签字日期： 年月日 签字日期：年月日 中文摘要 地震数据反演的目的在于通过求解反射系数序列来反映地下介质的分布规律，这就需从地震数据中可靠地消除地震子波的影响。然而受到有限带宽的影响，往往缺失反射系数序列中的低频和高频信息；同时由于受到不可避免的噪声的影响，反演得到的结果分辨率通常比较低。 本论文的目的在于使用迭代再加权的寻优方法由地震数据得到高分辨率的反射系数序列；该方法通过再加权的方法将稀疏优化问题转化为一系列权不断变化的线性方程组，在保证结果精度的前提下增加了求解的速度。对寻优过程中的线性方程组求解采用针对性的预条件化，并在权重的计算中引入伸缩变换，优化了整个寻优过程的稳定性和求解速度。经过大量的数值试验，得出了合理的参数选择范围，通过与真值的比较以及实际资料处理验证了本方法的实用性。 关键词：波阻抗反演，迭代再加权，Gauss-Siedel迭代，共轭梯度法，预条件化，伸缩变换 ABSTRACT The basic goal impedance inversion is to recover the layered structure of the surface as a function of depth from observed seismic data, which usually expressed as series of reflection coefficients. Such a goal could be achieved when the effects of seismic wavelet is eliminated in an effective way. However, it is hard to get high resolution reflection coefficients series because of the bandlimited effect of seismic data with the resulting loss of low and high frequency information about subsurface structure. The unavoidable noise reduced the resolution,too. The purpose of this dissertation is to invert a high resolution reflection coefficients series from seismic data using iterative reweighted technique. This technique transform the sparse optimization into a series of wighted linear function, speeds up the computing while keep signal-to-noise rate high. In order to stabilize and accelerate the solving,pre-conditioning according to the linear function is use