A Distributed-Acoustic-Sensing-Based Method for Early Warning of Urban Road Subsurface Cavities Using Vehicle-Induced Vibrations

Abstract

Rapid underground development in cities increases the hidden risks of roadbed cavities. Traditional geophysical methods, such as ground-penetrating radar (GPR), are costly and unsuitable for automated large-scale monitoring. This paper proposes a distributed-acoustic-sensing (DAS)-based algorithm for urban road cavity monitoring using vehicle-induced vibration signals. By leveraging routine urban traffic as a natural excitation source and DAS as the sensing medium, automated and continuous perception of subsurface-cavity hazards is achieved. A time–frequency joint vehicle-signal extraction algorithm is introduced to reliably isolate vehicle responses under complex noise conditions, and a physically interpretable Average Feature Ratio Indicator (AFRI) is constructed to identify long-term distribution shifts associated with structural anomalies. The AFRI remains highly stable within a narrow 0.70–1.00 range on cavity-free roads, even under low-level false positives (~1%). Experimental validation further shows that anomalous responses extend approximately 5 m from the cavity center. These findings confirm both the feasibility of the proposed method and its potential for practical early warning of subsurface cavities in urban road infrastructure.