首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于T-matrix的非线性参数估计方法
引用本文:王本锋,吴如山,陈小宏,陆文凯.基于T-matrix的非线性参数估计方法[J].地球物理学报,2016,59(6):2257-2265.
作者姓名:王本锋  吴如山  陈小宏  陆文凯
作者单位:1. 清华大学自动化系, 智能技术与系统国家重点实验室, 北京 100084;2. 中国石油大学(北京)油气资源与探测国家重点实验室, 北京 102249;3. Modeling and Imaging Laboratory, University of California, Santa Cruz, 95064, U.S.A.
基金项目:国家自然科学基金项目(U1262207),中国博士后科学基金(2016M590102)和国家科技重大专项课题(2016ZX05024001-005)联合资助.
摘    要:全波形反演可提供高精度的地下介质参数空间分布,但传统的全波形反演方法建立在Born近似的基础上,对初始模型具有一定的依赖性.为了摆脱Born近似的束缚,本文基于二维常密度声波方程,在De Wolf近似的前提下,借助传输矩阵(T-matrix)方法,深入研究了逆薄板传播算子(Inverse Thin-Slab Propagator,ITSP),实现了速度扰动的非线性估计.ITSP方法避免了Born级数方法在扰动较强、扰动区域较大时的发散性问题,且只经过一次扫描校正,计算效率较高.二维模拟数据分析验证了本文方法的可行性以及有效性.

关 键 词:全波形反演  De  Wolf近似  传输矩阵  逆薄板传播算子  Born级数  
收稿时间:2015-07-18

Non-linear parameter estimation method based on T-matrix
WANG Ben-Feng,WU Ru-Shan,CHEN Xiao-Hong,LU Wen-Kai.Non-linear parameter estimation method based on T-matrix[J].Chinese Journal of Geophysics,2016,59(6):2257-2265.
Authors:WANG Ben-Feng  WU Ru-Shan  CHEN Xiao-Hong  LU Wen-Kai
Institution:1. State Key Laboratory of Intelligent Technology and Systems, Department of Automation, Tsinghua University, Beijing 100084, China;2. State Key Laboratory of Petroleum Resources & Prospecting, China University of Petroleum, Beijing 102249, China;3. Modeling and Imaging Laboratory, University of California, Santa Cruz, 95064, U.S.A.
Abstract:With the development of seismic exploration and exploitation, it is necessary to study the accurate parameter distribution finely describing the subsurface. Full Waveform Inversion (FWI) can help us achieve this goal, however, traditional FWI theory is based on the Born approximation and the performance is dependent on the initial model. This is one of the main factors which prevent FWI being used widely in practice. In order to study the accurate parameter distribution, we studied nonlinear estimation method based on T-matrix which overcomes the initial model dependence.#br#The Inverse Thin-Slab Propagator (ITSP) is studied in detail both theoretically and numerically based on Transmission Matrix (T-matrix) and De Wolf approximation using the 2D acoustic wave equation with constant density, which can overcome the limitations of Born approximation. The ITSP method considers all orders of scattering effects and has no initial model dependence, besides, it is a regularized method in each correction step and has no divergence effects. One correction sweep (half from upper part, half from lower part) is involved and is computationally efficient. Based on the ITSP method, the velocity perturbation can be achieved and the velocity distribution can be obtained finally with the background velocity.#br#We designed three numerical models including ellipse ball with positive velocity perturbations, Gaussian ball with negative velocity perturbations and ellipse ball with both positive and negative velocity perturbations. With the ITSP method, we achieve the velocity distribution based on the known T-matrix and the T-matrix estimation method will be left for future research. The reconstructed results are consistent with the true velocity distributions and the relative errors are minor enough for these three models, which demonstrate the validity of the proposed method. Using the ITSP method, only one correction sweep is involved to achieve the velocity distribution, and it is time efficient. Therefore, this nonlinear estimation method may have wide applications in the future.#br#The accurate velocity distribution is achieved based on the known T-matrix using the ITSP method. 2D numerical examples of the designed models including different perturbation types, demonstrate the validity of the proposed method. The ITSP method considers all scattering effects and overcomes the initial model dependence of Born approximation, besides, it is a regularized method in each correction step and overcomes the divergence effect of Born series for strong velocity perturbation or large perturbation volume. It is also time efficient because only one correction sweep is involved (half from upper part, half from lower part). This method is based on the known T-matrix, suitable for smooth media, then T-matrix estimation method and the improved ITSP method for complex media will be developed in future work.
Keywords:Full waveform inversion  De Wolf approximation  Transmission matrix (T-matrix)  Inverse Thin-Slab Propagator (ITSP)  Born series
本文献已被 CNKI 等数据库收录!
点击此处可从《地球物理学报》浏览原始摘要信息
点击此处可从《地球物理学报》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号