Research On Wind Turbine Blade Damage Monitoring Technology Based On Audio Characteristic Parameters

Wind turbines are often located in harsh environments such as the seaside,mountains.Wind turbine blades are vulnerable to environmental erosion,resulting in wear,cracks,sand holes and other damages,which directly affect the smooth operation of the whole machine and power generation quality.In terms of blade damage monitoring,there is no relatively mature technology at home and abroad,and the staff often go through the way of "listening" and "seeing" Troubleshooting,high labor cost,low efficiency.In view of the above situation,this paper discriminates the damage of the wind turbine blade by monitoring the aerodynamic audio signal when the blade rotates,and designs a set of network monitoring system,the main research contents are as follows:(1)A set of ultra remote network transmission audio acquisition equipment based on TCP / IP protocol is designed.When the fan is running,the aerodynamic audio noise of the blade is collected and stored in the external storage equipment of the lower computer in the common audio file format.After the acquisition is completed,the 4G network is used to transmit to the monitoring room through the cloud service end,and the upper computer of the monitoring room analyzes and processes the signal.(2)According to the communication mode between the monitoring unit and the monitoring terminal of the wind turbine,a set of one point to many point networking communication scheme is designed,and the reliability of the communication is verified.(3)In order to detect and monitor blade damage effectively,the generation mechanism of blade aerodynamic audio signal and sound propagation mode are analyzed to determine the best sound acquisition position,and an improved spectrum subtraction method is proposed to denoise the signal.(4)In the aspect of blade damage identification,for the collected blade aerodynamic signal,in the time domain,from the perspective of signal similarity,the collected audio signal is compared with "transverse" and "longitudinal".The so-called "transverse" means that the similarity between different periods of the same signal is measured during acquisition,so as to filter the impact of random noise on the site.The so-called "longitudinal" is to be collected every day Based on the principle of dynamic time warping(DTW),the similarity of time-domain characteristics of different audio signals is calculated,and the damage of fan blades is preliminarily judged.(5)In frequency domain,ITD is used to decompose the aerodynamic audio signals of damaged blades.The characteristic parameters are extracted from the signals of each frequency band,the eigenvectors are constructed,and input to SVM for training and recognition.The results show that the algorithm reduces the calculation amount of feature extraction and greatly reduces the recognition rate compared with the traditional EMD decomposition Compared with the advantages of EMD decomposition,it has a higher recognition rate for damage types.