Research On Wind Turbine Blades Damage Identification Based On Harmonic Wavelet And Support Vector Machine

Blades are the key components of the wind turbine, the study of damage to which getsmore and more attention of the researcher in this field. Because of its large structure, irregularshape, complex material layers, and long-term work in hostile environments, so the problemwhich need to solve is how to monitor the health of wind turbine. Now monitoring methodwhich is used usually is determine damage situation of the blades by monitoring its modal,But the disadvantage of this method is low sensitivity, and has been unable to effectivelysolve. In order to solve this problem, this paper puts forward detecting damage situation ofwind turbine blades by the acoustic emission technology, and identify two types of injurypatterns of blades by using Support Vector Machine.Because of it can cause an internal strain of material when the blades is subjected toexternal damage, and will produce acoustic emission signal, it can realize the identification ofthe source of acoustic emission signals by collecting and analysising acoustic emission signals.First, turn on acoustic emission sensor, signal amplifier, data acquisition card and computer,to build acoustic emission signal acquisition experimental platform, fix acoustic emissionsensors on the blades by using a coupling agent. Then damage the individual static bladesmanually, simulate two types of acoustic emission signals which include crack propagationand edge damage, and acquire the acoustic emission when the blades is damaged.After collect the signal, decompose these two types of acoustic emission signals forfour-lever using harmonic wavelet packet and db10wavelet packet, and calculate energyvalues of each frequency band, after normalize process the energy value, these data is thefeature vector, train and study the feature vectors by using support vector machine, build themodel of damage identification. When identify the damage, the results of two wavelet packetfeature extraction are compared, the results show that using HWP and SVM combinedmethod can get good results. The method can identify different types of damage effectively, itcan help to identify early damage, blades can get timely maintenance and prevent furtherexpansion of damage.