Compressive Sensing and On-Board Data Recovery for Vibration–Based SHM

A primary challenge in the design of reliable and long–lasting Structural Health Monitoring (SHM) systems consists in ensuring real–time functionalities through cost–effective solutions. As such, energy–aware architectures demand the joint optimization of data sampling rates, on–board storage requirements, and communication data payloads. These requirements became particularly crucial with the development of mesoscale SHM systems, where the periodic gathering of signals from increasingly denser sensor networks made the data management task a primary issue. In the specific context of vibration–based SHM, where structural responses exhibit peculiar spectral profiles characterized by a sparse frequency content concentrated around the natural frequencies, the Compressive Sensing theory inspired compelling approaches for data collection and gathering to central processing units. The current work combines such advanced sub–Nyquist sampling procedures with a low-cost/low-power miniaturized Smart Sensor Network targeted on the extraction of vibration signals. The network is constituted by several recording nodes equipped with MEMS accelerometers and microcontrollers which are arranged in clusters, and microprocessors-based cluster heads in charge of data decompression and feature extraction for the characterization of the structural integrity.