基于波动方程表面多次波压制方法研究

发布时间：2018-03-13 01:01

本文选题：表面多次波　切入点：3D　出处：《东北石油大学》2015年硕士论文　论文类型：学位论文

【摘要】：多次反射波广泛存在于地震数据中,海洋资料尤为严重。目前,绝大多数的处理方法仅利用一次反射波进行成像,多次反射波的存在会严重掩盖一次反射波信息,多次波处理不当将直接影响后续的地震成像和解释,因此,地震资料多次波压制方法研究意义甚大。基于波动方程的表面多次波压制方法,又称SRME(Surface-Related Multiple Elimination),是一种数据驱动的方法,无需其他先验信息,运用了反馈迭代的思想,可有效压制复杂介质表面多次波。该方法由多次波预测和自适应相减两个关键步骤组成。目前,受计算和数据存储等因素影响,多次波预测大多局限于二维,或者通过诸多的假设、振幅匹配等将三维地震数据利用二维算法近似计算,其有悖于地下介质的真实情况,往往难以获取理想效果。常规基于L2范数匹配滤波方法尽管计算速度快,求解稳定,但是需要满足两个假设条件:一是压制多次波后记录能量最小;二是一次波与多次波具有正交性。本文在完成常规二维SRME预测方法和L2范数匹配滤波方法的基础上,创新性的完成了五方面内容:一是三维SRME方法预测多次波,该方法突破了二维算法的局限,充分考虑到地震波在地下介质中的空间传播效应,使得预测出的多次波更加精确;二是无需插值计算的基于抛物稀疏反演的三维多次波预测方法,该方法利用抛物Radon变换代替crossline方向多次波贡献道集的求和,避免了地震数据在crosline方向的内插和外推,同时减少空间假频的引入;三是基于L1范数匹配滤波方法,该方法能够避免基于L2范数滤波方法的两个假设条件,利用L1范数的鲁棒性对大值的异常保持稳健;四是基于L1/L2范数结合匹配滤波方法,考虑到L1范数和L2范数匹配滤波方法的各自特性,在较短的时间内得到收敛的维纳滤波器,使预测的多次波模型和原始记录中的多次波能够更好的拟合;五是将GPU(Graphics Processing Unit)和CPU(Central Processing Unit)协同并行加速技术成功应用于上述算法,极大地提高了多次波预测和自适应相减的计算效率。
[Abstract]:Multiple reflection waves are widely used in seismic data, especially in marine data. At present, most of the processing methods use only one reflection wave to image, and the existence of multiple reflection waves will cover up the information of one reflection wave seriously. The improper processing of multiple waves will directly affect the subsequent seismic imaging and interpretation. Therefore, the study of seismic data multiple suppression method is of great significance. The surface multiple suppression method based on wave equation, also known as SRME(Surface-Related Multiple delay, is a data-driven method. Without other prior information, the feedback iteration method is used to suppress the complex media surface multiple effectively. The method consists of two key steps: multiple prediction and adaptive subtraction. At present, it is affected by the factors such as computation and data storage, etc. The prediction of multiple waves is mostly confined to two dimensions, or through many assumptions and amplitude matching, 3D seismic data are approximately calculated by two-dimensional algorithm, which is contrary to the real situation of underground media. It is difficult to obtain ideal results. Although the conventional L2-norm matched filtering method is fast and stable, it needs to satisfy two hypotheses: first, the minimum recording energy is obtained after suppression of multiple waves; The second is the orthogonality between primary wave and multiple wave. On the basis of the conventional 2-D SRME prediction method and L2-norm matched filtering method, five aspects have been innovatively completed in this paper: first, the 3D SRME method is used to predict multiple waves. This method breaks through the limitation of two-dimensional algorithm and fully takes into account the spatial propagation effect of seismic waves in underground media, which makes the prediction of multiple waves more accurate. The second is the 3D multiples prediction method based on parabolic sparse inversion, which uses parabolic Radon transform instead of the sum of crossline direction multiple contribution traces, thus avoiding the interpolation and extrapolation of seismic data in the crosline direction. The third is based on L1 norm matched filtering method, which can avoid the two hypotheses based on L2 norm filtering method, and keep robust to the anomaly of large value by using the robustness of L1 norm. Fourth, based on L _ 1 / L _ 2 norm combined with matched filtering method, considering the characteristics of L1 norm and L _ 2 norm matching filtering method, we obtain convergent Wiener filter in a relatively short time. The predicted multiples model and the original records can be fitted better. Fifth, the GPU(Graphics Processing Unit and CPU(Central Processing Unit are successfully applied to the above algorithms. The computational efficiency of multiple prediction and adaptive subtraction is greatly improved.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.4

【参考文献】