Application of artificial intelligence in geophysical anomaly analysis from electromagnetic data

Abstract

Electromagnetic field data plays an extremely important role in representing geological structures and anomalous objects that exist from shallow depths to several tens of kilometers. The application of Artificial Intelligence (AI) in the analysis workflow of such geophysical data can bring practical benefits, such as providing fast and accurate results while saving expert resources. In this study, we focus on building a complete low-frequency electromagnetic (magnetotelluric) dataset by supplementing missing survey data and identifying underground anomalous objects from high-frequency electromagnetic data (Ground Penetrating Radar) based on diffraction mechanisms. The general AI model is developed based on interconnected neural network layers, including different networks such as MLP and CNN, to implement tasks of constructing a complete electromagnetic field dataset for the Olympic Dam mineral area in Australia, and distinguishing subsurface scattering objects from Ground Penetrating Radar data in Dong Nai province, Vietnam.