Intelligent geological interpretation of AMT data based on machine learning

AMT (Audio Magnetotelluric) is widely used for obtaining geological settings related to sandstone-type Uranium deposits, such as the range of buried sand body and the top boundary of baserock. However, these geological settings are hard to interpret via survey sections without conducting geological interpretation, which highly relies on experience and cognition. On the other hand, with the development of 3D technology, artificial geological interpretation shows low efficiency and reliability. In this paper, a machine learning model constructed using U-net was used for the geological interpretation of AMT data in the Naren-Yihegaole area. To train the model, a training dataset was built based on simulated data from random models. The issue of insufficient data samples has been addressed. In the prediction stage, sand bodies and baserock were delineated from the inversion resistivity images. The comparison between two interpretations, one by machine learning method, showed high consistency with the artificial one, but with better time-saving. It indicates that this technology is more individualized and effective than the traditional way.