Description
For existing bridges, reassessment and prognostic on lifespan is a key issue, especially in Europe where bridge assets are over 40 years old. Structural Health Monitoring (SHM) systems using various type of sensors, optical fibers, strain gauges and accelerometers offer a significant tool for safety verification approaches. In most of the use cases, engineering expertise is necessary to completely traduce measures as mechanical indicators and then be useful for reassessment. Moreover, without experimental reference solution, or knowledge of threshold values not to be exceeded, modal parameters or the axial deformation given by strain gauges or the vertical oscillations, an approach only oriented on signal processing can be quite limited. Data Mining (DM) offers then an alternative tool for damage identification of structures and identification of extreme loading effects. Recently, new techniques emerged, making instrumentation much easier and inspections cheaper. Time-frequency analysis tools such as the continuous wavelet transform have been recognized for several years for their ability to process vibration response signals, and to precisely identify the modal parameters of a structure. The topic of this research concerns the proposition of indicators using all data collected from various sensors (accelerometers, strain gages, temperature sensors) during few months on an existing bridge. Such indicators should be representative enough to allow an automatic structural health monitoring to detect abnormal load, and to assess damage using data mining techniques.
