Nonlinear Inversion For High Density Resistivity Method Based On NSGA-BP Neural Network Algorithm

Coal enterprises play an extremely important role in the development of national economy and society. At present, with the coal mining intensity and depth of our country strengthening and increasing continuously and there being more and more large coal fields, coal mine geology disaster has also became more and more severe. Geological prospecting, as the foundation of coalmine safety production, plays an important role in the prevention and control of coal mine geological hazard. High density resistivity method is one of the effective resistivity prospecting methods, and it is widely used in mining geological exploration.In explaining data obtained by high density electrical method, linear inversion represented by the least square method is still dominant.Because this method has low high inversion accuracy in the practical use,more and more scholars have started to devote themselves to the study on non-linear inversion, and BP neural network-based inversion of high density electrical method becomes a relatively active branch in the non-linear inversion. Based on the disadvantage of BP neutral network that its convergence is slow due to random initialization of weight and threshold value and it is easy to be trapped in the local minimum as well as the characteristic of neutral network nodes that the lower the order of magnitude of the weight, the stronger the network generalization ability,multi-objective optimization algorithm (NSGA-II algorithm) was adopted for combined calculation with BP neutral network algorithm and the training mean square error MSE of BP neutral network and the root-mean-square value of the parameters of the hidden layer were both taken as objective functions to carry out multi-objective optimization on BP neutral network, hence improving BP neutral network's accuracy in inverting the data obtained by high density electrical method.This paper introduced the inversion method and the detailed process of high density resistivity method based on the bp neural network through the computer models. The inversion results showed that NSGA-? algorithm could optimize the weight and threshold value of BP neutral network effectively, improve the global optimization performance of BP algorithm and the generalization ability of the neutral network and provide more accurate inversion results than traditional non-linear inversion algorithms and single BP neutral network algorithm.Finally, the coal mines grouting-drillings were taken as the object of study, measurement was carried out by high density electrical method and NSGA-BP algorithm was applied to invert the actually measured data. The result shows that the proposed method can be applied to explain the observed data effectively, and supplies the reliable basis for the engineering evaluation.