Damage detection through continuous monitoring of the response of a cable stayed bridge to temperature variations

One of the current challenges in the context of structural health monitoring (SHM) of important infrastructures, such as large span bridges, is how to manage the large amount of data produced by the SHM systems and make it readily available to the bridge management authorities in the form of performance indicators. In this study, the possibilities of early detection of structural damage through the analysis of the measured structural response of a cable-stayed bridge to daily and seasonal temperature variations are evaluated. A continuous monitoring system has been installed in the Corgo Bridge, here selected as case-study, which is acquiring data without significant interruptions for more than half year. The monitoring systems is described, and the software being used to access remotely and in real-time the measured data is presented. The ability of detecting damage using the structural response to thermal loads is evaluated applying Multilinear Regression Analysis (MLR) and Principal Component Analysis (PCA) for removing the environmental effects of the structural response. At this stage, simulated datasets were adopted so that at least one year of data can be used in order to remove the environmental effects from the structural response of the bridge in the undamaged state. The simulations were carefully performed by generating realistic thermal fields in the crosssections of the bridge using a finite element thermal analysis in which the boundary conditions are defined using the measured wind velocity, radiation, ambient temperature and air temperature inside the box-girder. The comparisons with the measurements of the existing temperature sensors (of which more than one year of data is available) reveal that a good agreement could be achieved. The thermal action is then applied to a mechanical finite element model of the bridge to obtain the simulated structural behaviour. Again, the comparison with the available measured data reveals a reasonable agreement, indicating that the simulated datasets are representative of the real structural behaviour. Several damage scenarios are simulated, mainly involving stiffness losses of the stay-cables and it is shown that the adopted methodology, jointly with the installed monitoring system, is able to provide early detection of small damages.