Scour is a major cause of bridge collapse worldwide. Conventional inspection techniques directly monitor the depth of soil near foundations. However, when scouring occurs, the holes can be filled with sediment, and so the scour cannot be detected by these methods. This paper presents an indirect, consequential method for an early and robust detection of bridge pier scour. Scour is detected using vibration analysis, by just monitoring the dynamic response of the structure. The method is based on sensitivity analysis of natural frequencies combined with inverse analysis and ambient vibration monitoring. Small variations in foundation stiffness are then estimated from small variations in measured frequencies. The bridge and its piers are modeled by beams, and the foundations by springs, to obtain an easy expression for the first order sensitivity of frequencies to foundation stiffnesses. Frequency variations are therefore expressed as linear combinations of the stiffness variations of the multiple foundations. This paper elaborates on a robust inversion algorithm that prevents from an amplification of the uncertainties in the variation of the measured frequencies. Based on a real bridge case study, the method is shown to detect scouring at early stage.

The case study refers to a wide class of typical motorway bridges in France, that is to say an 8-span bridge over the Loire River on the A85 motorway. The modelling assumptions of the bridge were validated through ambient vibration monitoring. Several scour scenarios were simulated by reducing the reference value of foundation stiffnesses. This study makes clear that out-of-plane modes only are sensitive to scour, while being independent of the type of support devices and of the temperature over a broad range of practical interest.

Foundation stiffness variations of 5% can be detected when the relative variation in out of plane frequencies is about 1%. This level of variation is much larger than the uncertainty associated with the measurement chain and the operational modal analysis performed by professional softwares.