Study On Rapid Forward And Inversion Of Large-Scale Gravity And Tensor Gravity Data Based On Multilevel Approach

In recent decades,with the rapid development of airborne gravity measurement technology,the fine inversion of large-scale,multi-area,and large amounts of gravity data and gravity gradient tensor component data has become possible.This leads to how calculating the fast forward modeling and inversion for large-scale gravity data has become an urgent problem to be solved.The forward and inverse problems can be regarded as a mathematical question,that is,solving differential equations,and the multilevel algorithm is the most effective method to solve this problem.Forward modeling is the basis for inversions.The computational efficiency of forward algorithms can determine the computational efficiency of inversion algorithms to a certain extent.To achieve fast,high accuracy forward,high efficiency and fine inversion and multiscale separation of multiple and weak anomaly sources,a translation equivalence technique is used.We then use the geometric trellis equivalent storage technique in the framework of multilevel algorithm,which can greatly reduce physical memory consumption and computation.Also,the idea of algebraic multigrid and multi-scale wavelet is introduced and combined.The multi-scale information of abnormal source can be obtained by means of the transfer operators between different levels of kernel coefficient matrices in basic framework of a multilevel method and multi-resolution properties in the wavelet.The rapid updating of multi-scale kernel function,the transfer operators of kernel coefficient matrices at different scales,and the inversion based on the framework of multilevel method are taken as the research objectives.Next,the fast forward and multi-scale inversion for large-scale gravity gradient tensor components data is investigated in this thesis.Finally,to verify the validity and reliability of the proposed algorithm,the fast forward modeling of the established models is tested.Compared to the point mass method(Li’s code)and the equivalent geometric trellis technique(Yao’s code),the numerical results show that the proposed method can greatly reduce memory consumption and speed up computation.Furthermore,both multilevel and Occam inversion methods are applied to the real gravity data from Geoscience BC of Canada.The results prove the effectiveness of the proposed method.