Analysing technics with optical fibre networks : assessment of the accuracy of vibration measurements with fibre optics

This report presents the results of Task 3.3.3, within European Rail Joint Undertaking (ERJU). In the Horizon Europe program. This task evaluates the possible application of fibre optic (FO) for the control of environmental vibrations caused by pass-bys of trains. Within this task Deltares assessed the accuracy of fibre optic (FO) vibration measurements relative to conventional vibration measurements by accelerometers. The work is based on combined datasets from Holten (2024) and Culemborg (2020), where both FO and accelerometer measurements were collected during train pass-bys. The findings demonstrate that FO measurements reproduce low-frequency vibrations (≤ 20–25 Hz) with good reliability. Characteristic bogie-related frequencies (~11–15 Hz) are consistently observed in both FO and accelerometer data. Above 25 Hz, FO signals show strong attenuation under the tested gauge lengths, with high-frequency features such as sleeper-passing peaks (~50–60 Hz) absent from the FO response. It remains to be determined whether alternative interrogator configurations or smaller gauge lengths can recover this higher-frequency content. The practical implication is that FO and accelerometers are complementary measurement techniques. FO offers dense and continuous spatial coverage of low-frequency vibrations along the railway, while accelerometers provide detailed high-frequency information at discrete locations. Building on these results, the next phase of the work will focus on machine learning (ML) models to translate FO data into conventional vibration metrics. The first objective will be to predict peak ground velocity (PGV) from FO signals using gradient-boosted decision trees (XGBoost). PGV is a robust descriptor of soil vibration in a continuum medium and provides a practical target for initial model development. Additional information such as train type, speed, soil conditions, and sensor distance will be included as model features. If successful, this approach will demonstrate the feasibility of extending existing dark fibre networks into a scalable monitoring tool for railway vibration assessment, aligning with the broader ambition of innovative and cost-effective infrastructure monitoring across the European rail network.