Research On Airborne Time-domain Electromagnetic Data Preprocessing

Airborne Electromagnetic(AEM)is a geophysical exploration method which use airplane as the carrier of the electromagnetic instrument. It has been widely used in mineral survey, geological mapping, hydrography, environment inspection and so on.This paper has discussed kinds of techniques of airborne time domain electromagnetic data preprocessing supported by the 863 plan'Time-domain helicopter-borne airborne electromagnetic survey theory research and system design'of the'Airborne geophysical prospecting techniques system'. The main research work and conclusions are as following:Firstly, we analyzed the features of the airborne time domain electromagnetic data and the common noises in the Airborne Electromagnetic such as sferics, coil motion noise and other kinds of EM noise. The natural noise spectral in the 5Hz-100kHz range is primarily due to sferic which is caused by lightning primarily. Coil motion noise is of a different character which is primary between 1Hz-1kHz, produced by the movement of the receiver coil in the Earth's magnetic field when the magnetic flux changing inside the receiving coil. As the restriction of the stacking times in AEM survey, effective noise removal is essential to obtain high quality data.Secondly, we simulated the airborne time domain electromagnetic data and different kinds of noises. And the effectiveness of different ways for removing different noises have been discussed. For sferics, pruning is a useful method to reduce the noise level as the percentage of the number of data affected by recognized sferic events is not very high. For coil motion noise, we used high-pass filter, wavelet analysis method and polynomial method respectively to remove the motion noise. A high-pass filter with the cut-off frequency of 20Hz has been designed to filter the airborne time domain electromagnetic data with coil motion noise. Meanwhile the data has been decomposed into 13 levels through wavelet transform techniques. And the low-frequency segment of the 13th level was reduced as motion noise. Polynomial method combined with Lagrange method has been employed with a good job. We use a 3 degree polynomial to represent the motion noise for each half cycle, and translate finding unknown coefficients of polynomial into minimal problem via Lagrange method according to the characteristics of AEM data. For EM noise in off-time data, Butterworth low-pass filter, data composite filter, wavelet analysis method were adopted respectively. So we compared different wavelet basis, decomposition levels and thresholds for wavelet analysis method. The conclusion is that wavelet analysis method with Sym8 wavelet basis, 5-scale decomposition and default threshold is a good choice for this kind of noise. And it could get a high signal-to noise ratio.Thirdly, averaging window techniques and sampling techniques have been discussed in this paper. For averaging window techniques, traditional square averaging window technique and new trapezoid averaging window have been used respectively. We can see that new trapezoid averaging window could raise the data utilization ratio and the noise level at late time decreased. For sampling techniques, original data are sampled at the interval of logarithm, approximate logarithm and overlapped channels respectively. The conclusion is that data sampled at overlapped channels could obtain better resolution of the sounding geotectonic resistivity structures.Finally, we convert the airborne time domain electromagnetic datas which are preprocessed by different methods into Conductivity Depth Imaging based on Neural Network. Through the results of the CDI, we can see that effective techniques of airborne time domain electromagnetic data preprocessing is significant for the inversion and interpretation of the airborne time domain electromagnetic data.