基于多尺度高阶奇异谱熵的信号特征提取方法

提出了基于变分模态分解(VMD)的高阶奇异谱熵的特征提取方法,并应用在滚动轴承故障诊断中。首先,使用4阶累积量切片代替奇异谱熵分析(SSEA)的协方差矩阵,引入VMD分解实现方法多尺度化,提出信号多分辨高阶奇异谱熵分析(M-HSSEA)方法;通过信号分析,VMD解决了模态混叠的问题,且能够实现信号滤波,同时该方法提取的熵特征向量增强了相空间重构参数鲁棒性;通过和小波奇异谱提取特征的方法对比,结果表明所提出的方法在克服频率混叠现象,提取的特征点总体离散度小等方面更具优势;最后,结合深度信念网络分类器实现了对故障的分类,实验结果验证了所提方法的有效性和可行性。

Signal Feature Extraction Method Based on Multi-scale High-order Singular Spectrum Entropy

A feature extraction method based on variational mode decomposition (VMD) for high-order singular spectral entropy is proposed and applied to fault diagnosis of rolling bearings. Firstly, the fourth-order cumulant slice is used to replace the covariance matrix of singular spectrum entropy analysis (SSEA), and the VMD decomposition method is introduced to multi-scale. The multi-resolution high-order singular spectrum entropy analysis of bearing vibration signal is proposed. Through signal analysis, VMD solves the problem of modal aliasing and can realize signal filtering. At the same time, the entropy feature vector extracted by the method enhances the robustness of phase space reconstruction parameters. By comparing with the wavelet singular spectrum extraction feature, the results show that the proposed method is more advantageous in overcoming the frequency aliasing phenomenon and the small overall dispersion of feature points. Finally, the classification of faults is realized by combining the deep belief network classifier. The validity and feasibility of the proposed method are verified by the results.