Damage identification in plate-like structures using frequency-coupled l1-based sparse estimation

This article proposes a practical and effective damage identification approach for plate-like structures. This approach measures the back scattering caused by damage, and decomposes it into individual contributions from each defect, using the responses of the healthy structure as a dictionary. A data-driven model is used, which circumvents the challenge of numerically simulating the effect of damage, yet does not require training data from known-damaged structures. The decomposition itself is performed using sparsity-promoting optimization, which reduces the number of required measurements and streamlines the inspection procedure. A novel frequency-coupled method is proposed to obtain the desired spatial sparsity of the estimated damage, which results in improved accuracy compared to the previously proposed frequency-decoupled method. Damage identification is demonstrated on a 600 mm x 600 mm composite plate, using a single accelerometer and 7 impact hammer hits. The performance is evaluated on 6 damage scenarios, for 7 accelerometer positions, and for SNRs ranging from 30 to 0 dB. Detection and localization are shown to be excellent up to 5 defects and down to 15 dB SNR, and to remain robust and predictable outside of that range. These results are compared to reference methods and a significant improvement is observed.