Adaptive signal decomposition and dispersion removal based on the matching pursuit algorithm using dispersion-based dictionary for enhancing damage imaging

Abstract

This paper aims to develop a method for high-resolution damage imaging for a sparsely distributed sensor network on a plate-like structure. Techniques for dispersion removal and signal decomposition are indispensable to accurate damage localization. By combining the dispersion-removed wave packets with the damage-imaging algorithm, a point-like damage can be precisely localized. In this article, a matching pursuit algorithm was utilized to decompose overlapping wave packets and then recompress the dispersion. The matching pursuit dictionary was constructed based on an asymptotic solution of the dispersion relation for Lamb waves in toneburst wave packets. The dispersion-based Hanning-window dictionary provided the parametric information for the extracted wave packets, such as propagation time-delay, dispersion extent, and phase. The parameters were leveraged for the dispersion-removal algorithm. Results of the simulation indicate that the proposed algorithm is capable of recompressing multiple dispersive wave packets with the different modes. Finally, the proposed approach was validated by the results of the experiment using a sparse array of piezoelectric wafers on an aluminum plate. Extracting the parameters of individual wave packets and removing the dispersion through matching pursuit, the algorithm for minimum-variance imaging produced a high-quality image with a fine spatial resolution. The image artifacts were significantly suppressed, and the accuracy was improved by 62.1% compared to conventional minimum-variance imaging.