针对CSEM三维有限元正演中普遍存在的局部人工源的奇异性、起伏地形难以处理、计算量庞大等问题，传统的基于结构化网格、全局加密策略的算法效率、计算精度差强人意。为此，申请者拟引入区域分解矢量有限元算法，该算法将复杂、大型的CSEM三维正问题拆分为若干个简单子域，在相邻区域的分界面上施加Robin传输条件，且根据各子域上求解目标体局部特性分别求解。这种具有高度并行度的矢量有限元算法，有效解决节点有限元存在的“伪解”问题，为超大规模三维CSEM的MPI并行求解提供了一条有效的途径。再结合非结构化网格、后验误差估计的自适应策略、面向目标体自适应加密等整套思路，整体提升CSEM三维正演精度与效率，力求以最小计算量求解测点位置上电磁场值，达到快速、高精度求解复杂三维CSEM问题的目的。

The algorithm efficiency and calculation accuracy of traditional methods of secondary field calculation and global encryption strategy is barely satisfactory in processing the undulating terrain, and the high gradient and singularity of local artificial sources that are found prevalent in 3D CSEM (Controlled Source Electromagnetic) FEM (Finite Element Modeling) direct modeling. Therefore, by introducing the algorithm on domain decomposition vector finite element, this project divides the large and complex CSEM 3D forward problems into several simple subfields, applies Robin transmission conditions on the interface between adjacent domains and hence solves the problems respectively according to the properties of the solving objects in the subfields. Such vector finite element algorithm of high parallelism can effectively solve “pseudo solution” in the node finite elements and provide a valid path to the MPI parallel solution of super-scale 3D CSEM. When being combined with the complete thinking which includes the adaptive strategy of unstructured grid and posteriori error estimation and the object-oriented adaptive refinement strategy, it can improve the whole accuracy and efficiency of CSEM 3D forward modeling and acquire the electromagnetic value at the measuring point position with the smallest amount of calculation, so as to realize the goal solving the complex 3D CSEM problem with the least amount of computation in a fast and accurate way.