Sparse Denoising of Eddy Current Signals from Conductive Material Defects Based on K-SVD Dictionary Learning

Abstract

Eddy current testing technology is widely used in the defect detection of conductive material and the integrity assessment of key components. Because the eddy current signal is easily affected by the complex electromagnetic environment during the testing process, it causes a lot of noise in the signals, which affects the analysis of eddy current signals. Based on the sparse and redundant representation of eddy current signals, a dictionary learning method based on K-means singular value decomposition (K-SVD) for adaptive sparse representation is studied. The noisy signals are sparsely decomposed on the redundant dictionary constructed by learning, and then the sparse representation vector is reconstructed, and the noise is processed into the residual to be discarded in the reconstruction process, thereby achieving the separation of the signal and the noise to achieve the denoising effect of the eddy current signal. Finally, the experimental results verify the accuracy and effectiveness of the proposed method in eddy current signal denoising.