The Study Of Non-linear Inversion Theory Of TEM Sounding

Transient electromagnetic sounding method in recent years has developed rapidly athome and abroad, as a very potential geophysical exploration method, it gained massivepopularity by the people. With the further research of the theory, how to improveinterpretation accuracy of the field observed data ulteriorly, and reduce the likelihood ofmultiplicity of solutions, it has been the direction of the continuous efforts for geophysicalworkers. At present the most widely used inversion method is linear approximation ofnonlinear problem, by neglecting the high order terms in the iteration to accomplishobjective function optimization, this method has a lot of problems: inversion processheavily relies the initial model, losing too much useful information, extremely easilyfalling into local extremum in solving process, and inversion process needs to muchmatrix operations, not only time-consuming but inversion result is not stable.As a more effective geophysical inversion method, when the nonlinear inversionappeared, it became geophysical workers′research hotspot, its principle is through byvarious methods to solve nonlinear problem directly, to achieve the mapping from dataspace to model space, planning the inverse problem to be a solution of a functionalextreme value problem, and then using all kinds of optimization methods to solve theproblem, the method can quickly get the parameters for the model when solving nonlineargeophysical inverse problems, and make the field observed data get more preciseexplanation.This paper is based on the modeling theory of transient electromagnetic soundingmethod, at the same time, according to the characteristics of field observed data, weestablished the objective function of the nonlinear inverse problem, in view of the existingnonlinear inversion method is easily falling into local extremum and over learning, it is thefirst time introduced a kind of very novel and excellent intelligent optimization algorithmin current optimization field, artificial bee colony algorithm, to solve inverse problem. Thealgorithm has many obvious advantages when the solution evaluation is nonlinear andmulti-extreme, it does not need to know the special information of prepared solvingproblem, what we need to do is comparing the quality, through by the searching behaviorof every single artificial bee, we will make the global optimal value is highlightedeventually, and it has quickly convergence speed. Through by the detailed calculating ofthe algorithm parameters at the Matlab workbench, finally we determined a set of best inversion parameters, and then we verified with the geoelectrical model of two, three, fourlayers as an example. After a series of model optimization, based on Matlab platform, weworked out the nonlinear inversion software for transient electromagnetic soundingmethod, and applied it to the inversion of field observed data, through by the comparisonwith the borehole data and other inversion methods, and then we realized furtheroptimization of the inversion system, finally the evaluation for actual extension ability wasmade.Through the calculation of different geoelectrical model and effects of actualapplication we can find that, in the inversion calculation accuracy, speed, etc, artificial beecolony algorithm is much better than the traditional linear inversion method or otherexisting nonlinear inversion method, and the robustness of the algorithm is stronger, thealgorithm is also more stability. The research results of this topic is a new breakthrough inthe field of current geophysical inversion, although at present just stopping in1Dinversion, but our subsequent research work can be further expanded to2D,3D, it has abetter development prospect.