Adaptive Spectral Subtraction Noise Reduction Method for Generalized Acoustic Emission Signals of Fatigue Cracks in Wind Turbine Towers

As a novel means for crack detection in wind turbine towers, acoustic emission technology is often subject to noise interference during field applications, so appropriate methods are required to mitigate the impact of noise. Based on the characteristics of crack acoustic emission signals, a frequency-division adaptive spectral subtraction filter estimation method was proposed. This method expands the applicable conditions of the original spectral subtraction method, allowing for the noise reduction of crack acoustic emission signals in complex noise environments without the necessity of detecting effective signals. In this method, spectral subtraction filters for different frequency bands were obtained through dynamic mean estimation, and then weighted and summed to generate a new spectral subtraction filter. The noise reduction effect was validated using acoustic data collected from fatigue tests. Results demonstrate that this method is superior to wavelet thresholding method and traditional empirical mode decomposition (EMD) in terms of noise reduction. The accuracy of fatigue crack identification after noise reduction has been improved to 97.6%, aiding in the accurate identification of crack signals.