A Study on Deep Learning for Inverse Problems of the Electrical Resistivity Imaging

Geo-electrical resistivity imaging method is often used in studies such as engineering geology and hydrogeology, and is one of the most widely used exploration geophysical techniques. The artificial and natural electric field monitoring system established by our team in the past five years has the ability of collecting a large amount of electrical potential data in a short period of time and transferring data to a specified cloud drive via a real-time network. By receiving a large amount of geoelectrical potential data in real time, researchers can focus on the calculation of inverting electrical resistivity structures subsurface. However, the large amount of electrical data continuously and intensively transferring back to the laboratory also highlights the limitations of the traditional inversion procedures. Firstly, a large amount of data causes a burden on the memory resources in the traditional inverse calculation. Secondly, the intensive scanning results also highly require an increase in the inverse calculation speed. This project therefore proposes the introduction of a deep learning network to assist the traditional inversion process. It is expected that this integration process can provide faster and more accurate resistivity images to enhance the electrical resistivity imaging method in applications of hydrogeology and engineering geology.