Hardware–Oriented Data Recovery Algorithms for Compressed Sensing–Based Vibration Diagnostics

The huge amount of data to be processed is still a challenge for the current Structural Heath Monitoring (SHM) sensor networks and a primary reason for the development of hardware–oriented signal processing techniques. In the specific case of vibration–based inspections, the sparse spectral distribution of structures in dynamic regime have made the Compressed Sensing (CS) paradigm a compelling solution. In this work, the on–sensor deployment of data recovery procedures is tackled from an edge–computing perspective, aiming at selecting the most performing strategy which allows for the joint optimization of implied memory storage, latency and consistency of the retrieved structural information. An experimental campaign conducted on a steel beam undergoing ground motion excitation revealed that the Orthogonal Matching Pursuit (OMP) strategy might be a promising candidate for sensor deployment, since it attains the highest reconstruction levels while minimizing the associated memory/time cost.