Research On Fusion Interpretation Method Of Airborne Transient Electromagnetic Method And Magnetotelluric Method

Airborne transient electromagnetic method is one of the commonly used exploration methods at present.How to combine aviation transient electromagnetic method and ground electrical method to obtain better survey results is an important issue that geophysical workers need to face.This article is based on the BP neural network and principal component analysis as the basis of the national key research and development plan of "A typical coverage area airborne geophysical technology demonstration and processing interpretation software platform development" major project " Research on Airborne Geophysical Data Comprehensive Processing Method and Software Development "(No.2017YFC0602202),start multi-source data fusion and interpretation research,and realize data fusion and interpretation between airborne transient electromagnetic method and magnetotelluric method.In this paper,based on the basic principles of electromagnetic method,the one-dimensional forward response expressions of airborne transient electromagnetic method and magnetotelluric method are derived,establish a variety of two-layer and three-layer geoelectric models,the forward modeling program is used to obtain the airborne electromagnetic and magnetotelluric method forward modeling responses and calculate the apparent resistivity.Then the principal components of the apparent resistivity of the two methods are analyzed and the first principal component is taken;Taking the first principal component as output,the apparent resistivity of the two methods is the intput,form a training sample set and normalizes it.According to the basic theory of artificial neural network,combined with the idea of data fusion,the network topology and training parameters are determined,and a single hidden layer BP neural network that meets the requirements is established and trained.Use the trained network to fuse the apparent resistivity of the two typical electromagnetic models of two typical electrical models,and use the normalized parameters of the network input information to denormalize the network output to obtain the pseudo apparent resistivity,and the comparison of the fictitious apparent resistivity curve after fusion and the apparent resistivity curves of the two electromagnetic methods are compared and analyzed.The results show that the pseudo-resistivity curve obtained by the fusion of the BP neural network can correctly reflect the resistivity characteristics of the model,and can also weaken the false anomaly in the magnetotelluric method,verifying the BP neural network's ability effectiveness of fusion with magnetotelluric apparent resistivity data;Then use the pseudo-resistivity as input,and the geoelectric model parameters as the output to form a new training sample set,build a new double-hidden layer BP neural network and train it,use the trained BP neural network to interpret the pseudo-resistivity data,then,the interpretation results are compared with the geoelectric model.From the results,the interpretation results are in good agreement with the resistivity and thickness of the model,which further verified the feasibility and practicability of the method.Finally,the trained network is used for fusion processing and interpretation of the measured data,and good results are obtained,which verifies that the method has practical value.