3-D gravity and magnetic inversion for physical properties using stochastic subspaces
重磁异常三维物性反演随机子域法方法技术

Focused on the existing difficulties and problems in 3-D inversion for physical properties,the paper brings forward stochastic subspace methodology for 3-D inversion for physical properties.Firstly,it computes separately and saves the geometric trellis which keeps unchanged during forward simulation and inversion to avoid repetitive computation so as to increase the speed of forward simulation and inversion computation.Secondly,it uses symmetry to realize equivalent computation,which distinctly lowers the requirement of trellis computation and storage.And thirdly with stochastic subspace inversion method it reduces the number of dimensions of inversion.In addition,it controls the distribution of the subspaces generated through probability method to realize the new mechanism of constraint.The computations of model and field data demonstrate the effect of the methodology which is hopeful to be of practicality-oriented 3-D inversion for physical properties of large scale gravity and magnetic data and meets the requirement of explanation of 3-D inversion.