The analysis of short signal segments and its application to Drive-by bridge inspection

‘Drive-By’ damage detection is the concept of using sensors on a passing vehicle to detect damage in a bridge. At highway speeds, the vehicle spends a short amount of time on the bridge: it may not even go through a full oscillation, resulting in only a partial signal of the bridge motion being detected. Given that the spectral resolution of standard signal processing techniques depends on the length of data in the signal, they cannot be used to identify the bridge frequency accurately. In addition, the nonlinear and non-stationary nature of the vehicle-bridge interaction system poses challenges. An optimisation approach is proposed here as an alternative to standard signal processing techniques to overcome the challenges of short signals and the nonlinear nature of the drive-by system. Signal pollution due to the road profile is overcome using time-shifted bridge curvatures, a novel damage indicator.