Structural health monitoring in the Sydney harbour bridge using spectral moments

Abstract

© 2017 International Society for Structural Health Monitoring of Intelligent Infrastrucure. All rights reserved. The motivation behind this paper is to develop a spectral-based damage identification scheme using output only acceleration responses. The presented method is in the context of non-model-based damage identification methods and does not require any representative numerical/analytical model of the structure. The method utilizes spectral moments of the response as damage sensitive feature. Spectral moments directly retrieve information from the power spectral density of the response. Unlike the modal data that only provide information at a limited number of eigen-frequencies, spectral moments capture information from the entire spectra, hence they can distinguish any subtle differences between a normal and distorted signal. The feasibility of the approach in damage identification was validated using real data from the Sydney Harbour Bridge. There are approximately 800 jack arches over a total distance of 1.2 km need to be continuously monitored. For this study, two instrumented jack arches were considered. These joints are located on the eastern side of the bridge underneath the bus lane near the north pylon. One of these two joints had a known crack in 2012, along the front face propagating toward the surface of the deck, while the other joint was intact. This damage was repaired in 2013. Acceleration data were collected from tri-axial accelerometers mounted on the base of each joint before and after repair. The presented spectral-based method along with the hypothesis testing involving the KS-test were applied to obtain a decision on whether or not the structure is damaged. Spectral moments with different orders were also investigated. It was demonstrated that the proposed spectrum-driven feature can reliably distinguish between the healthy and damaged joints which is of great importance for the asset owner. The presented results illustrated high potential of this approach to identify damage in real-life structures.