Study On Airborne Transient Electromagnetic Quasi-Resistivity Imaging Method Based On Artificial Neural Network

The ATEM(Airborne Transient Electromagnetic Method)belongs to the time domain aeronautical electromagnetic system and is one of the commonly used methods for airborne geophysical exploration.Based on the theory of time domain airborne transient electromagnetic method,and combined the national key research and development plan of "Deep Resource Exploration and Mining" project "A typical coverage area airborne geophysical technology demonstration and processing interpretation software platform development" major project " Airborne Transient Electromagnetic The research on data processing methods and software development of law and terrestrial electrical methods,this paper analyse the characteristics of airborne transient electromagnetic data,and realized the artificial transient neural resistivity imaging based on artificial neural network.In this paper,the airborne transient electromagnetic 1D forward modelling theory is systematically discussed,and a variety of uniform half-space models are established.The airborne transient electromagnetic response is obtained by forward modelling,and formed a forward sample set.By analysing the airborne transient electromagnetic response curve,the flight height is used together with the time constant and the response amplitude as the mapping parameters of the quasi-resistivity.Based on the basic theory of artificial neural network and combined the characteristics of pod-type airborne transient electromagnetic data,the training sample set and network model parameters are determined,and a double hidden layer BP neural network conforming to the mapping relationship between quasi-resistivity and airborne transient electromagnetic response is established.By analysing the mean square error curve,the training end criterion of BP neural network is summarized.The airborne transient electromagnetic response of a variety of typical theoretical geoelectric models was imaged using the trained BP neural network,and compared with the results of the numerical simulation algorithm of the aeronautical transient electromagnetic field based on the translation algorithm.The results show that BP neural network quasi-resistivity imaging is superior to the all-time apparent resistivity numerical method.The BP neural network method has a better approximation to the low-resistance layer and is closer to the resistivity of the real model;the response to the high-resistance layer is slightly better than the all-time apparent resistivity numerical method.The applicability and effectiveness of artificial neural network to aeronautical transient electromagnetic imaging problems are proved.Finally,the measured data of the pod-type airborne transient electromagnetic field were processed,and the quasiresistivity-depth section calculated by the BP neural network was obtained,which is consistent with the actual geological conditions.The research has certain theoretical significance and practical application value.