Study And Application Of CSAMT Inversion Based On The Genetic Algorithm

The Control Source Audio-frequency Magnetotelluric (CSAMT) is developedwith the Magnetotelluric (MT), because of that's artifical electromagnetism sourceand high quality data, becoming an important non-earthquake method in geophysicsexplore. Inversion is one of the key taches in CSAMT. Aiming at traditional method'sdepending on initial model and having no sensitive strata, the Genetic Algorithms(GA)are applied in CSAMT inversion in this article. By analyzing the theory of GA, someimproved methods are putted forward there and the programs of GA base on theVisual Fortran environment have been developed.Genetic Algorithms(GA) are random searching methods similar to naturalselection and natural genetic mechanism, which are advanced by professorJ.H.Holland in 1975. It is simple, general, quite robust and fitting for parallelprocessing, and has been successful applied in widely fields in the past 20 years. GAhas such good qualities as no differential coefficient requiring, no local linearization,low request for initial model and so on. The key of genetic algorithm is efficiency.Because of the faintness of the parameters, we should extend the bound value of theparameters. Otherwise the forward calculating of CSAMT is very complex and weneed the loop of the forward calculating between the inversions. So the speed of GAprogram is limited. But the forward calculating of MT is simpler than CSAMT, so wecan use the inversion of MT as an assistant method to create a grope of initial model.The search area of layer parameters can be confirmed with the modeling error of theoptimistic model in initial model. Then the inversion in CSAMT can be executed baseon the initial model and new search area and the speed have been improved.The GA inversion reflects the electronic structure more than the linear way bythe experiment in the data inversion for the Shihuashu tunnel of railway fromKunming to Hekou of Yunnan province.