项目名称: 地球物理数据的正则化与稀疏优化反演方法研究
项目编号: No.11271349
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 王彦飞
作者单位: 中国科学院地质与地球物理研究所
项目金额: 50万元
中文摘要: 本项目重点研究地球物理反演问题中的地震数据和重磁数据的正则化处理方法。从数学上看,这些地球物理数据的正则化反演问题均可以归结为由第一类有限秩算子方程导出的非线性稀疏优化问题,但问题的规模十分庞大。利用非线性稀疏优化技巧处理地球物理数据的正则化反演问题,具有十分重要的现实意义。本项目拟针对实际问题从物理建模、算法设计和最终的计算机实现进行系统的优化反演方法研究。我们将利用几类问题的结构特性,选择合适的先验信息,建立地球物理数据正则化的稀疏反演模型,克服反演计算的非唯一性;在算法实现上,具体研究带稀疏约束的子空间信赖域算法和交替方向算子分裂方法,克服计算的不稳定性并提高反演效率问题。在理论模拟与实际数据验证相结合的基础上编写可以为实际应用服务的高效软件程序。
中文关键词: 数据正则化;稀疏优化;正则化反演;地球物理;
英文摘要: We study geophysical inversion in this project: seismic data regularization and gravimetric-magnetic data regularization methods. From mathematical point of view, geophysical data regularization problems reduce to the finite-rank operator equations of the first kind which requires solving a nonlinear sparse optimization problems. Hower, these problems are in large scale. Using nonlinear sparse optimization methods tackling geophysical data regularization problems is important in practice. In this project, we focus on the physical modeling of the practical problems, algorithmic designing and computer simulations using the optimization techniques. Based on the specific structure of the geophysical problems, we study choosing proper a priori information, establishing sparse inversion model for geophysical data regularization, overcoming the non-uniqueness in computation; in algorithmic realization, we study the subspace trust region method and the alternating directions operator-splitting method with sparse constraints,overcoming the instability during computation and improving the inversion efficiency. Based on theoretical simulations and practical data validation, we make computer software program which may be used in practical applications.
英文关键词: Data regularization;Sparse Optimization;Regularizing inversion;Geophysics;