风电机齿轮振动的小波降噪与风沙侵害分类

WAVELET DE-NOISING AND CLASSIFICATION FOR GEAR VIBRATION OF WIND TURBINE SUFFERING FROM WIND SAND EROSION

根据小波变换系数,优选db8小波基函数对风沙侵害电机齿轮振动信号进行小波降噪;采用分段阈值二值化分析信号的幅频关系,定义振动信号幅频特征矩阵;提出用振动特征矩阵与零特征矩阵间的欧氏距离对风沙侵害分类。搭建模拟风沙侵害实验平台,实验测试变转速和不同沙粒大小对振动的影响。实验表明风沙引起的齿轮振动主要集中在中低频段,且随转速增加而振幅变大;特征矩阵间的欧氏距离对电机转速不敏感,但随风沙颗粒增大而变大,随信号采样频率增加而变小。特征矩阵间的欧氏距离分类特征明显。

Based on the the gear vibration signal analysis of wavelet transform coefficients, the wavelet basis function db8 was selected to reduce ＇ s noise. Then the vibration signal was filtered by wavelet transform. According to the piecewise threshold method, the signal amplitude-frequency relationships were analyzed, and the signal high-dimensional feature matrix to describe the wind sand erosions grade was defined. Euclidean distance was adopted for the classification between the vibration signal feature matrix and the origin matrix. A wind sand erosion test platform was built, and the experiment to test the vibration effects under the different rotor speeds and sizes of sand was completed. The experiments showed the vibration produced by sand focused on the intermediate and low frequency, and signal amplitude increases with the speed. The experimental data showed the Euclidean distance increases with increasing the size of sand, declines with the sampling frequency increasing, and is not sensitive to the rotor speed. Euclidean distance between the feature matrixes is valid for the classification.