Acoustic detection of fan blade faults based on dynamic Cauchy swarm algorithm to optimize support vector machine

Fan blades operate in outdoor environments, where the detection of sound signals is susceptible to interference from background noise such as random loads, wind speed, rainwater, and other ambient noise. Therefore, this article proposes an acoustic detection method for wind turbine blade faults based on a dynamic Cauchy bee colony algorithm-optimized support vector machine. First, the signal is preprocessed using a Butterworth bandpass filter, and the full frequency band is divided into sub-bands using the octave band feature extraction method. Based on frequency domain analysis, the natural frequency offset of the blade is determined. Next, the dynamic Cauchy bee colony algorithm is applied to optimize support vector machine parameters, while moving average and bandpass filtering are used to smooth the noise power curve and extract impeller speed information. The experimental results show that the proposed method converges in fitness value after 22 iterations, with a detection time of only 6.8 seconds and small fluctuations in impeller speed amplitude. In terms of classification performance, the accuracy of detecting normal samples is 0.95, the recall rate is 0.96, and the F1 score is 0.95. The method demonstrates high prediction accuracy and stability for various types of fault samples and can be reliably applied to the acoustic detection of wind turbine blade faults.