EMI signal processing via compressive sensing

Taking account of the inefficiency in data transmission and storage caused by the enormous data collected during structural health monitoring (SHM), the compressive sensing (CS) technique was employed in data compression of the electromechanical impedance (EMI) signal in order to achieve efficient data transmission and storage. In this paper, the sparsity of EMI signal was analyzed by matching pursuit (MP), and the Gauss random matrix was taken as the measurement matrix, thus both the sparsity of the signal and the incoherence of the measurement matrix, as required by the CS, were met simultaneously. As an example, an one-dimensional damaged bar was used to perform the EMI analysis, in which the data compression efficiency and the noise resistance ability of CS were discussed by taking the root mean square deviation (RMSD) as the damage index. The research results indicate that, after the intervention of CS, the transmission bandwidth and storage space are only 28% of the original signal. The damage of 100 different trials can all be identified if the measurement number is four times of the sparsity. And when the signal-to-noise rate (SNR) is greater than 20 dB, the original signal can be reconstructed stably from the CS results. The research results prove that, as a data processing method, CS can be applied in the EMI data processing.