A method for inversion of aerial transient electromagnetic data based on lstm network

Abstract

The invention discloses an aviation transient electromagnetic inversion method based on LSTM network. The LSTM network is established by dimension, the input data is the aviation transient electromagnetic time domain signal, the sensor height is spliced ​​into the feature layer to participate in the calculation, and the output data is the resistivity model parameters; 3) According to the error reduction curve, select the appropriate training set size and training period, Ensure the training effect and training efficiency, and complete the network training; 4) Judge the generalization ability of the network according to the fitting of the network inversion results and its forward response; 5) Use the LSTM network to complete the inversion of aviation transient electromagnetic data and obtain the The resistivity distribution structure of the medium. The inversion method is stable and accurate, and has fast calculation speed, which can improve the interpretation efficiency of aviation transient electromagnetic data, and then provide technical support for real-time imaging of aviation transient electromagnetic data.

Description

technical field [0001] The invention provides a fast inversion method of earth electromagnetic data, in particular to an inversion method of aerial transient electromagnetic data based on LSTM network. Background technique [0002] Aerial transient electromagnetic method is an important shallow surface exploration method. It can adapt to complex terrain, has fast acquisition speed, wide detection coverage, and is highly sensitive to low-resistance anomalies. It is used in mineral exploration, groundwater monitoring and oil reservoir imaging. has been widely used. [0003] The aerial transient electromagnetic method has high sampling efficiency and large amount of observation data, and it continuously emits current to excite the secondary field during flight, which has multiple sources. Interpretation work presents enormous computational challenges. At present, the commonly used methods for interpretation of aviation transient electromagnetic data are mainly divided into de...

AI Technical Summary

Problems solved by technology

The aviation transient electromagnetic response data decays exponentially with time, which belongs to a typical time series. The above-mentioned neural network structure is difficult to fully extract the time series features; Boundary of combined resistivity anomalous body

[0005] In summary, the existing data interpretation methods for aviation transient electromagnetic response still have limitations, and it is necessary to study an efficient and accurate data interpretation method

Images