基于EWT和多尺度熵的轴承特征提取及分类

Feature Extraction and Classification of Bearings Based on EWT and Multi-Scale Entropy

轴承的故障信息提取直接决定了诊断的正确与否,为了能准确地识别轴承状态,提出了一种基于经验小波变换和多尺度熵的轴承特征信息提取及分类方法。该方法通过提取信号频域相邻最大值间的极小值,对Fourier谱进行自适应划分,并构造合适的小波滤波器组提取不同的模态;再引入多尺度熵,对最优模态建立的粗粒向量进行状态分类。试验分析表明:与EEMD相比,该方法具有更优的自适应特征提取和故障分类特性。

The correct rate of diagnosis is directly determined by fault information extraction of bearings. In order to ac- curately identify state of bearings, an extracting and classifying method for feature information of bearings is proposed based on empirical wavelet transform and multi - scale entropy. By extracting the minimum values between the adjacent maximum values of signal in frequency domain, the Fourier spectrum is divided adaptively, and a suiTab, wavelet filter bank is constructed to extract different modes. Moreover, the multi - scale entropy is introduced to classify the state of coarse grain vector based on optimal mode. The experimental analysis shows that the proposed method has better adap- tive feature extraction and fault classification feature than EEMD.