A comparison of compressive sensing based reconstruction algorithms used for data loss recovery in SHM applications - Simulation and experimental demonstration

Recently, the concept of Structural Health Monitoring (SHM) is simplified using output-only structural analysis with the help of Wireless Sensor Networks (WSNs). However, due to harsh environment and low power sensor nodes, data losses are a matter of concern in WSNs. To efficiently tackle these data losses, Compressive Sensing (CS) based recovery techniques have become popular for SHM. The main advantage of CS technique is that it transfers the computational load to the receiver (Base station) and saves a lot of transmitter power. However, reconstruction algorithms implemented at the receiver are computationally complex. In this paper, a comparison of existing popular reconstruction algorithms is performed. It is found that using Fixed Point Continuation with Active Set (FPC_AS) method is showing good performance for SHM applications. This algorithm is a modified version of l1 norm based optimization technique where extra knowledge of the signal is used for faster reconstruction. Along with simulation results, performance analysis using hardware (sensor node) deployed on a real bridge is also presented. It is observed that using FPC_AS algorithm, faster reconstruction of the signal is achieved with the reconstruction quality better than the other algorithms.