Application Of GB Decomposition Method With Genetic Algorithm For MT-Geological3D/2D Model

MT composite models reflect well the natural condition in which the main distortions to the MT data often come from very complex near-surface imhomogeneities, which is responsible for the most of problems in analysis. Research on MT impedance tensor distortion and decomposition theory is meaningful in MT data processing, especially in inversion.Nowadays GB decomposition method named after Groom and Bailey is used the most widely. It is based on the feature that2D structure with a local3D electric-inhomogeneity covering. The method decomposes measured impedance into site gain, three distortion matrices and regional impedance tensor. As a result, a sequence of over-determined equation systems can be obtained. Traditional GB theory uses local optimization algorithm, such as sequential quadratic programming (SQP), generalized inverse method, etal, to solve the equations after transforming non-linear problem into the linear one. It is hard to calculate partial derivative matrix when lack of prior information in initial values selection, and easy to fall into local extremum and other weak points make traditional linear methods unpractical.In order to solve the defects in GB theory mentioned above, we introduce a kind of global optimization algorithm called Genetic Algorithm (GA) to GB decomposition in this paper. Firstly, we prove that Swift rotation is not effective on reducing regional impedance tensor and spindle azimuth, and raise the viewpoint that the solution is periodical besides the phase difference (π/2) between the regional trend and tropesis even when there is no inhomogeneity exist. Secondly, using the improved method to process classical synthetic data from Alan G. Jones and Gary McNeice and the forward data of a layered model with near-surface3D anomalous body. The result shows that GA in GB theory is effective. Finally, a series of practical measured MT data are processed and we get the apparent resistivity and phase curves and relative distortion factors, which represent the underground electric characteristics more reasonably compared with predecessors’ results. All above steps help prove that GB decomposition method with Genetic Algorithm in this paper is useful and practical.