Noise Tomography Based Urban Ground Collapse Monitoring Using Distributed Acoustic Sensing

Abstract

With extensive development of urban underground space, ground collapses occasionally occur, resulting in property damage, injuries, or fatalities. Therefore, establishing an effective ground monitoring and collapse prevention system is necessary for reducing collapsing risk. Distributed Acoustic Sensing (DAS) technology has great potential for ground collapse monitoring around rail system for capacity of long-distance high-density monitoring and immunity to electromagnetic interference (EMI). In this study, experimental investigations were conducted using a high-performance DAS interrogator integrated with a 4 km-long distributed optical fiber to record background ambient noise inside a subway tunnel. Noise2Noise (N2N) unsupervised denoising model was employed to preprocess the DAS data across different time spans, thereby enhancing the signal-to-noise ratio (SNR). Additionally, noise tomography method in terms of passive multichannel analysis of surface waves (MASW) algorithm was utilized to extract the surface waves and calculate the dispersion curves containing soil layer information Stacking methods, including Phase-Weighted Stacks (PWS) and Generalized Average of Signals (GAS), were applied to further improve the SNR of the dispersion curves. Ultimately, a 2D S-wave velocity model to 45m depth was constructed, revealing low-velocity soil defects such as hollow and settlement. The result demonstrates that integrating DAS with tomographic imaging using urban ambient noise can provide a reliable basis for preventing ground collapse hazards.