Research on Conventional ERT Inversion and Improved AMT Inversion Based on Deep Learning Denoising Data in Tunnel and Road Detection: A Case Study in Sichuan Province

Abstract

The safety of tunnels and roads is crucial for traffic safety. Due to the presence of adverse geological features, which can cause serious problems in tunnels and on roads, there is an urgent need for comprehensive geophysical investigations to determine their distribution. This will provide additional technical information to help ensure the safety of engineering projects. This study uses the tunnels and roads in Jiulong County, Sichuan Province, as an example. Integrated geophysical methods were employed to arrange ERT sections at tunnel entrances and exits, as well as on geophysical slopes. The primary focus was on the thickness of the overlying layers and the geological conditions of the rock and soil within a certain depth range above the tunnel design line. For the longitudinal cross-section of the main tunnel, Audio Magnetotelluric (AMT) were primarily used to investigate fault zones, karst formations, aquifers and rock mass grades. Combining electromagnetic data denoising and inversion using a U-Net neural network with ERT inversion clearly revealed the underground geological conditions of the tunnel. The tunnel alignment is characterised by a mixture of stable quartz-rich rock masses and metamorphic rock masses. Overall, the deep rock mass of the tunnel axis is stable, with fractures in some transverse sections. The intersection is relatively stable, interspersed with weathered and fractured zones. These findings provide valuable insights into advancing geophysical techniques for investigating tunnel sites under complex geological conditions.