Continuous SHM of Railway Bridges Based on Vibration Analysis of Qualitative, Selected, Asynchronous Data

Since mid-2022, a consortium of 3 partners, SISGEO, LEMTA and SNCF Réseau, is carrying out the AUDACE project, as part of the “Smart Bridges” project in France. Our objective is to provide bridge managers with real-time information regarding the occurrence of a collision of a railway bridge deck. This type of incident, likely to cause damages that can impact the stability of the structure, represents more than 50% of incidents recorded on the SNCF Réseau (SNCFR) bridges. SNCFR selected several bridges regularly experiencing this type of incident for SISGEO to install a set of relevant sensors. Among them, the AD-SIGNUM solution provides a continuous and global SHM of the bridge. AD-SIGNUM is based on several key technologies. First, geophones are used to measure vibrations of structures giving access to low amplitude signals in either quiet or noisy environment. Then, a particular effort is paid on the processing of asynchronous signals in order to make deployment and maintenance of autonomous sensors easy, flexible, and efficient. Finally, rather than gathering an endless amount of data, we focus on a qualitative perspective. This in-depth approach aims to measure the own vibration of the structure, and its evolution over time, apart from punctual non-typical disruptions. The LEMTA will compare different “Machine Learning” algorithms with supervised, unsupervised and reinforcement learning approaches to select the best performing approach to identify the critical parameters to detect anomalies and signs of impact damage to finally draw conclusions for the bridges manager to take the right decisions.