Bridge Modal Identification Using Crowdsourced Vibration Data from Passing Vehicles

Despite rapid advances over the past four decades, sensing technology is still not an integral part of vibration-based bridge inspection protocols due to the high cost and scalability issues of dedicated sensor networks. In recent years, indirect monitoring through vehicles as sensors has emerged as an alternative that addresses some of the challenges of dedicated sensor networks. Our work leverages existing mobility networks of smartphone sensors in moving vehicles to measure the spatiotemporal vibrations of bridges cost-effectively and at unprecedented scales. Although the advantages are clear and significant, this novel paradigm poses its own challenges. The signals acquired from inside the vehicles are not purely the bridge vibration. Instead, they are the vehicle’s dynamic response when traversing over a bridge corrupted with noise as well as the impact of the road pavement surface. This renders the bridge modal identification problem extremely challenging based on data collected from a single vehicle passing over a bridge. To address these issues, we propose a time-frequency analysis-based method for the estimation of bridge modal properties from an ensemble of vehicle datasets. The proposed approach has been tested both numerically and through field experiments. In this paper, we report our findings, demonstrate the approach’s efficacy, and discuss the existing challenges that still require an interdisciplinary research effort. The complete development of a holistic, crowdsourced, data-based modal identification scheme will revolutionize vibration-based bridge monitoring and help enhance the road transportation infrastructure’s resilience.