基于多尺度解调谱熵的轴承故障特征提取方法

Research on Bearing Fault Feature Extraction Method Based on Multi-scale Demodulation Spectrum Entropy

为了解决直驱风力发电机主轴轴承故障诊断问题,针对实际工程中振动信号的复杂特性,提出了一种基于改进经验模态分解(EMD)的多尺度解调谱熵的特征提取算法。多尺度解调谱熵利用EMD自适应分解特性与信息熵融合,首先对轴承振动信号进行EMD分解,将得到的各阶固有模态函数(IMF)分量进行Teager能量算子解调,获得不同频段的解调信号;其次,对各解调信号构造能量矩阵,并进行奇异值分解求取解调谱熵作为特征向量,从而实现对信号的多分辨率分析;最后,通过支持向量机(SVM)对实例数据进行故障分类实验,实现了较高的分类准确率,证明了该方法对于轴承故障诊断的有效性。

According to the fault diagnosis of spindle bearing in Direct-driven Wind Turbine,Combing with the complex characteristics of vibration signal in practical engineering,we proposed a signal analysis and fault diagnosis method based on multi scale demodulation spectral entropy. Multi scale demodulation spectral entropy using the EMD decomposition characteristics and information entropy fusion. Firstly,EMD was used to decompose the bearing vibration signal contains a finite number of intrinsic mode functions,then use Teager energy operator demodulation get demodulation signal of different frequency bands,and then to the demodulation signal matrix structure mode,and calculating the multi-scale demodulation spectrum entropy value as a feature vector,so as to realize the multi-resolution analysis of the signal. By using the support vector machine to classify the sample data,It is proved that the method is effective for the fault diagnosis of bearing.