Three-Dimensional Modeling And Inversion Of Controlled Source Audio-frequency Magneto Telluric Based On Domain Decomposition

CSAMT, developed by MT, is an artificial-source electromagnetic method. There are anti-interference and high signal-to-noise ratio about CSAMT. It is sensitive to various depth of underground by different frequencies. It also has a well resolution in vertical direction. Because of these advantages, CSAMT is widely used in not only prospecting of mineral and water resources but also investigating of oil and gas energy.Additionally, CSAMT plays an important role in geological survey, geothermal prospecting, and environmental engineering investigating.There are three numerical method simulating and inversing CSAMT, like finite element method (FEM), finite differential method (FDM) and integral equation method(IEM). FEM is very accurate in calculating, while it demands quite high for supplying boundary conditions. IEM is a high efficient method by only dividing abnormal area,nevertheless much time cost on the computing integral of dyadic Green's function both of electrical and magnetic field. FDM is too dependent on mesh division, bad of which may make the result be divergence. The divergence correction is required in the program of FDM. Considering both advantage and disadvantage of these methods, the GILD method is proposed. Firstly, the research zone is separating into two part — one is boundary zone simulated by IEM, and the other is inner zone simulated by FEM.IEM has no requirement on boundary condition, which is helpful for FEM simulating.Similarly, the stiffness matrix and load matrix is fast calculated in FEM, which can take the place of integral of dyadic Green's function in IEM. Thus, joint two methods can save the computing time and improve the efficiency of three-dimensional modeling of CSAMT. It also improves the calculation accuracy and greatly reduces the requirement for the storage. Moreover, owing to independence of each subdomain, the parallel computing, like OpenMP, is suitable for algorithm of GILD.Damped least square method is used for the inversion of CSAMT in this dissertation. In the process of inversion, updating the damped factor can get the best inversion model that makes fitting residual become least and is smooth. Inversing the same observe data in different size of research domain by GILD get the similar result,which conclude that the range of research domain is not limiting condition for inversion.