Mode Separation of Complicated Guided Wave in Plate-Like Structures Based on Sparse Bayesian Learning Approach

Abstract

Guided wave based methods has been demonstrated to be the effective techniques for damage detection in plate, tube and pipe etc. structures during the last two decades. However, the nature of guided waves, such as multi-mode, dispersion effect and mode conversion, results in the difficulty of guided wave packets recognition. This work uses a special piezoelectric wafer to generate both Lamb waves and guided shear horizontal (SH) waves in plate-like structures. The latter is employed to detect the damage taking advantage of its non-dispersive properties. However, the SH wave packets without distortion are still disturbed by the other mode wave packets, such as S0 and A0 mode of Lamb waves at low frequencies. In this study, based on a robust sparse Bayesian learning (RSBL) technique, a novel mode separation method is proposed to identify the SH wave packets. An over-complete dictionary is first designed by using the propagating waveforms with various distances. The dictionary is then employed to decompose the received signals containing SH waves and Lamb waves by the RSBL algorithm. The sparseness of wave packet is utilized and the posterior mean vector of the retained basis vectors in the dictionary can be used to determine the propagation distance of wave packet. Furthermore, the posterior uncertainties give a measure of the inference confidence. A numerical study is performed to validate the proposed method and the results show that the proposed method is capable to separate the SH mode wave packets with high accuracy.