Research On Feature Extraction And Fault Detection Of Acoustic Signals Of Wind Turbine Blades

Compared with photovoltaic power generation and nuclear power generation,wind power generation has a shorter installation cycle and flexible installation scale,so that the installed capacity of wind turbines has been continuously increased.People are paying more and more attention to the production and maintenance of various components of wind turbines.As the core component of wind turbine energy conversion,wind turbine blades are prone to safety hazards such as delamination or blisters under the constantly changing stress coupling.If the blades cannot be inspected in time,personnel and economic losses will be caused.Therefore,it is necessary to detect blade damage in time.At present,most wind farms use manual line inspections to understand the operating status of the blades.However,in the face of the increasing scale of power plants,manual line inspections are very time-consuming and inefficient.In order to improve the detection efficiency of wind turbine blades,a variety of detection methods have been developed.This paper focuses on the extraction of acoustic features of wind turbine blades and the remote detection of blade damage,mainly from three aspects:(1)Development of wind turbine blade acoustic signal acquisition and transmission system.By determining the best position and angle to collect the blade acoustic signal,the microphone and MAX9814 microphone amplifier are used to collect the fan blade acoustic signal,the PCM1808 digital-to-analog converter converts the acoustic signal into binary data,and the STM32F103C8T6 single-chip microcomputer is equipped with a GPRS module to transmit the data to the PC.On the PC side,restore the data to WAV format audio files on the PC side.(2)Based on LabVIEW blade acoustic signal feature extraction.Using Lab VIEW software to preprocess the acoustic signal and octave band processing,carry out frequency domain conversion on the blade acoustic signal,and select the appropriate filter and filter order to intercept the blade acoustic signal.Through 1/6 octave processing,the frequency band power of the acoustic signal of the normal blade and the faulty blade is extracted,and the feature extraction of the acoustic signal of the wind turbine blade is realized.(3)Establishing a fault detection model for wind turbine blades.Based on SVM classifier and BP neural network in MATLAB,blade fault detection models are established respectively,and the SVM classifier and BP neural network are trained using the extracted acoustic signal power characteristics to determine whether the blade is faulty and the fault situation.The test results show that both the SVM classifier and the BP neural network can complete the detection of wind turbine blade faults,and the accuracy of the BP neural network is higher than that of the SVM classifier.