Wind turbine blade fault diagnosis method based on beamforming and improved perceptual linear prediction algorithm

Abstract Aiming at the problem that wind turbine blades are prone to failures and it is difficult to identify fault features in aerodynamic noise signals under complex operating environments, this paper proposes a wind turbine blade fault diagnosis method that combines beamforming technology and an improved PLP algorithm. By applying the beamforming theory, the strategy of wind turbine blade airspace division and microphone array signal spatial filtering is constructed, which enhances the intensity of the target direction signal and suppresses the background noise. In order to improve the feature extraction ability of blade aerodynamic noise signal, the perceptual linear prediction (PLP) algorithm is improved, and the extraction efficiency of the algorithm on the fault features in the aerodynamic noise signal of wind turbine blades is enhanced by reconfiguring the mapping relationship between the physical frequency and the Bark frequency, and adjusting the distribution of equal loudness weights. Through the K-means clustering analysis with the removal of wind noise clusters, the effective differentiation of blade sound signals with different health states is achieved, and the location of fault occurrence is determined based on the clustering results of the signals in each airspace. Finally, the effectiveness and robustness of the proposed method are verified based on an experimental platform, which provides effective technical support for the health monitoring of wind turbine blades.