Vulnerability assessment of existing bridges to scour, based on an indirect monitoring approach and machine learning tools

Abstract

Detecting scour in railway bridges is possible by locating accelerometers and GPS on carriages of passing trains and processing the resulting signals. This research aims to detect scour based on these drive-by measurements, obtained from na instrumented passing vehicle. Signals from multiple train passages will be collected before and after scour repair to determine the change in bridge behavior. Measurements from a train in the UK passing over the Carlisle Bridge will be provided through In2Track3, an ongoing Horizon 2020 project. In the first stage of the numerical approach, off-bridge conditions are considered. The carriage vibrational responses to track with different ground conditions – represented by altering the stiffnesses in a Winkler spring model – are calculated. In second stage, the bridge ‘apparent profile’(AP), which is made up of the true profile on the bridge plus components of bridge/track deflection, will be computed. The Moving Reference Influence Line, i.e., deflection per unit load at a moving reference point, is found from the measured deflections. Bridge support stiffnesses will be modified to represent the loss of stiffness due to scour. Then, signals from the instrumented in-service train carriage i.e., measured AP, will be processed. Finally, an optimization algorithm will find foundation stiffnesses by minimizing the sum of squared differences between the calculated AP and the corresponding measured AP. The presence of scour will be determined by the difference between the stiffness values in the scoured and repaired cases. The results will help to optimize retrofits or develop mitigation measures to scour.