Acceleration-based Bridge Scour Monitoring

This paper comes from a research project focused on the safety assessment of bridges using camera-based technologies. The project is developing methods that transform measured sensor signals and video images into a form that is highly damage-sensitive for bridge safety assessment. It will advance sensor-based structural health monitoring with computer-vision and accelerometer-based techniques, leading to practical applications for bridge damage detection.
Many sensor types have been used in test installations, with varying degrees of success. Strain gauges and transducers are well established technologies and sufficiently accurate sensors are available at a reasonable cost. However, strain transducers can only detect damage if it occurs close to the point of measurement and are completely insensitive to scour-induced settlement. Deflection at any point on a bridge is a function of support conditions and the flexural stiffness at all points. As such, it has the potential to provide an indication of damage at any point. Deflection can be difficult to measure and some of the partners in this project are working to develop image analysis techniques to improve the accuracy of camera-based deflection measurement systems. Doppler laser vibrometers measure the derivative of deflection with respect to time, i.e., velocity, but they are expensive and it is impractical to deploy large numbers on smaller bridges.
Acceleration is the 2nd derivative of deflection so it is, in theory at least, sensitive to scour damage. Furthermore, accelerometers are widely available and can provide accurate measurements at a reasonable cost. This paper reports on the use of acceleration measurements for bridge scour monitoring. Traffic induced acceleration on a bridge is the result of a range of excitations. The signal is influenced by a number of vehicle-related factors such as speed, inter-axle spacing and tyre and suspension properties. In this project, the portion of the acceleration signal in the region of the bridge first natural frequency is filtered from the raw input in order to amplify the portion of the signal most likely to be influenced by bridge damage. Vehicle/bridge dynamic interaction simulations are used to show the nature of the response, before and after filtering. It is shown that the filtered signal is considerably more sensitive to bridge damage than the original raw signal and has good potential for bridge health monitoring.