A denoising method for acoustic emission signals in wind turbine blade fatigue testing based on SVD-VMD

Abstract

To improve the accuracy and reliability of acoustic emission (AE) signal processing in wind turbine blade fatigue tests, this paper proposes a denoising method that combines singular value decomposition (SVD) and variational mode decomposition (VMD). The technique utilizes SVD to extract principal components, which are then used to determine the optimal number of VMD modes, effectively reducing mode mixing. To address the issue of mode selection, a combined metric based on permutation entropy (PE) and wavelet energy ratio is introduced to identify and eliminate noise-dominated modes, thereby achieving accurate extraction of effective modes. Comparative studies with traditional methods, such as ensemble empirical mode decomposition (EEMD) in both simulation and real fatigue test scenarios, demonstrate that the proposed approach outperforms in preserving signal features while suppressing noise. Multidimensional similarity analysis and envelope spectrum validation further confirm the effectiveness and practical value of the proposed method for the health monitoring of wind turbine blades.