Periodic Audio Signal Extraction of Wind Turbine Blades Based on Sound Pressure Level Determination

The audio signal caused by the damage of the blade is mixed with the periodic audio signal generated by the rotation of the wind turbine blade. Therefore, the key prerequisite for the blade damage detection is to accurately extract the periodic blade audio signal in the complex noise environment. This paper proposes the extraction method of the periodic blade audio signal based on the determination of sound pressure level. Firstly, the high-pass filter is used to reduce low-frequency noise interference. Secondly, the true peak sequence is constructed by the true peaks in the peaks extracted from the audio signal with the definition of the adaptive threshold, and then the sequence is improved by means of searching the missing peaks. Finally, the periodic blade signal is extracted by comparing the variance of the peak-to-peak interval, and the method for evaluating the quality of the periodic blade signal is proposed to find the signal segment that can accurately reflect the period of the blade audio signal. Because the sound pressure level is related to the peak-to-peak interval of the audio signal, the peak-to-peak interval parameters in the periodic blade signal extraction steps are set to accurately extract the periodic blade signal and calculate the period of the blade audio signal, on basis of the sound pressure level of the audio signal in the paper. The analysis of the actual data in the field shows that the proposed methods can effectively extract periodic blade signals, with the recall, the precision, and the F1-score all above 95%.