Denoising algorithm for shotcrete acoustic emission signals based on IPSO-VMD and its application in damage feature extraction

Abstract To address the challenges of nonlinearity, non-stationarity, and strong noise interference in acoustic emission (AE) signals during the damage evolution of shotcrete, this study proposes a novel adaptive denoising algorithm based on improved particle swarm optimization-variational mode decomposition (IPSO-VMD). By employing the minimum average envelope entropy as a fitness function, the optimal number of decomposition modes k and the penalty factor α in VMD are automatically determined, thereby eliminating the subjectivity inherent in conventional parameter selection. Effective intrinsic mode functions are identified using the correlation coefficient method, while noisy components are further refined through wavelet packet decomposition. The denoised AE signal is reconstructed by combining the retained effective components with the processed noisy components. Simulation experiments demonstrate that, compared with empirical mode decomposition and other commonly used denoising techniques, the proposed algorithm achieves substantial improvements in signal-to-noise ratio (SNR ⩾ 15.99 dB), root mean square error (RMSE ⩽ 0.0510), and normalized cross-correlation coefficient (NCC ⩾ 0.9877). Additional validation through uniaxial compression tests on shotcrete confirms that IPSO-VMD effectively suppresses environmental noise, reducing low-frequency noise events by 98.8%. Moreover, the dominant frequency characteristics are better preserved, with energy concentration in the 100–150 kHz range enhanced by 7.91%. Based on RA-AF features and Gaussian mixture model classification, the identification accuracy of tensile cracks improves by 5.2%. Overall, the proposed algorithm offers a reliable approach for the refined extraction of shotcrete damage evolution features, thereby supporting structural safety assessment and early warning applications.