Data-driven damage detection for bridges through vibration structural health monitoring

Bridge structures are essential for social and economic development, but they are costly to maintain. More efficient planning and better investment of resources are needed. To this end, Structural Health Monitoring (SHM) techniques are used to collect data on the behavior of the bridge in terms of engineering and environmental variables to support decision making. However, without predefined objectives for the monitoring campaign, it often results in large, unwieldy databases from which little or no value can be derived. To tackle this problem, this PDEng project focuses on a methodology that translates vibration global SHM data into a damage indicator. To this end, (i) two types of damage sensitive features obtained from the vibration data were thoroughly explored; (ii) a process based on Principal Component Analysis (PCA) was used to address the high dimensionality space of the data. Besides, an approach to calibrate the reference period based on the PCA was proposed; (iii) a one-class support vector machine to perform damage detection using damage sensitive features was implemented; and (iv) the validation was carried out based on two case studies, the Z24 bridge, and the SMC bridge benchmarks. In particular, for the latter, anomalies were detected with respect to the reference period four months before the closing of the bridge when the damage was found through on-site inspection.