Acoustic Pulse Extraction of Fan Blades based on Energy Threshold Optimization

Due to the influence of a complex environment, the accuracy of wind turbine blade damage detection is significantly affected by the pure noise segments between adjacent blade acoustic pulses in the audio signals of wind turbine blades collected on-site. An acoustic pulse extraction algorithm for fan blades based on energy threshold optimization is proposed. Firstly, signal preprocessing based on wavelet packet de-noise is applied to improve the periodicity of the audio signal. Subsequently, the preprocessed signal is segmented into frames and windowed, and an energy threshold function is proposed based on the energy analysis of audio frames. Then, a firefly algorithm with a dynamic step is proposed to optimize the energy threshold, thereby completing the search for the central frame sequence of the blade acoustic pulse. Finally, the central frame sequence is used to realize the extraction of the blade acoustic pulse. Experimental results demonstrate that the blade acoustic pulses can be effectively extracted by the proposed algorithm to significantly reduce interference from pure noise segments.