Research on signal denoising algorithm based on ICEEMDAN eddy current detection

Abstract This study addresses the challenges of non-stationarity and significant background noise interference in eddy current detection signals by proposing a noise reduction method based on Improved Complete Ensemble Empirical Mode Decomposition with Adapted Noise (ICEEMDAN). The process commences with the signal being decomposed using Improved Complete Ensemble Empirical Mode Decomposition with Adapted Noise into a finite number of Intrinsic Mode Functions (IMFs). Each Intrinsic Mode Function is then evaluated for the presence of high-frequency noise using a Power Spectral Density (PSD) analysis. The high-frequency noise present in the Intrinsic Mode Functions is then reduced using Normalized Least Mean Squares (NLMS) before being reconstructed with the remaining Intrinsic Mode Functions. Subsequently, the reconstructed signals are subjected to another round of decomposition using Improved Complete Ensemble Empirical Mode Decomposition with Adapted Noise. The Pearson Product Moment Correlation Coefficient (PPMCC) is utilised to calculate the correlation between the Intrinsic Mode Functions within each layer, retaining those with a strong correlation to further attenuate noise. Ultimately, the local maxima judgement method selectively amplifies defect signals by assessing changes in peak and valley degrees, thereby improving the signal-to-noise ratio of the eddy current detection signal. The experimental results demonstrate that, in comparison to the use of only the conventional Improved Complete Ensemble Empirical Mode Decomposition with Adapted Noise and Normalized Least Mean Squares denoising methods, the proposed method increases the Signal-to-Noise Ratio (SNR) by 1.08 dB and 2.31 dB, respectively, and decreases the Mean Square Error (MSE) by 106.9 and 223.9, respectively. The false alarm rate for stainless steel welded tubes with defects is 1.4%, while the false alarm rate for stainless steel welded tubes without defects is 0.4%.