Abstract: Multi-scale reverse dispersion entropy(MRDE） might effectively measure the complexity of time series, but MRDE had defects in coarse-grained structure and lacked the effective use of other channel information in characterizing the nonlinear fault characteristics of rolling bearings. To accurately extract fault features from bearing signals, combined with the ideas of refinement and generalized composite multi-scale, the multi-variate sample entropy theory that characterized the multi-variate complexity of synchronized multi-channel data was applied to the rolling bearing fault diagnosis and RGCMvMRDE was proposed. Then, a rolling bearing fault diagnosis method was proposed based on RGCMvMRDE and gravitational search algorithm support vector machine(GSA-SVM). Firstly, the RGCMvMRDE was applied to comprehensively characterize the fault feature information of rolling bearings and the fault feature sets were contracted. Secondly, GSA-SVM was used to identify the fault type intelligently. Finally, the proposed fault diagnosis method was applied to analyze experimental data of rolling bearing with comparing with the existing fault feature extraction methods based on MRDE, generalized MRDE(GMRDE） and refined composite multi-yariate multi-scale permutation entropy RCMvMPE. The results indicate that the proposed method may effectively and accurately identify different fault types and fault degrees of rolling bearings and the diagnosis effectiveness is better than those of the compared methods.

Key words: refined generalized composite multi-variate multi-scale reverse dispersion entropy(RGCMvMRDE), rolling bearing, fault diagnosis, feature extraction

摘要： 多尺度反向散布熵能够有效度量时间序列的复杂性，但在粗粒化构造上存在缺陷，且在表征滚动轴承非线性故障特征时缺乏对其他通道同步信息的有效利用。为了准确提取轴承信号的故障特征，结合精细化和广义复合多尺度的思想，将表征同步多通道数据多变量复杂度的多变量熵理论应用到轴承故障诊断中，提出了精细广义复合多元多尺度反向散布熵（RGCMvMRDE）。在此基础上，提出了一种基于RGCMvMRDE与引力搜索算法优化支持向量机（GSA-SVM）的滚动轴承故障诊断方法。首先，利用RGCMvMRDE全面表征滚动轴承故障特征信息，构建故障特征集；其次，采用GSA-SVM对故障类型进行智能识别；最后，将所提方法应用于滚动轴承实验数据分析，并将其与现有基于多尺度反向散布熵、广义多尺度反向散布熵和精细复合多元多尺度排列熵的故障特征提取方法进行了对比。研究结果表明，所提RGCMvMRDE不仅能够有效和精准地诊断轴承的不同故障类型和故障程度，且诊断效果优于上述对比方法。

关键词: 精细广义复合多元多尺度反向散布熵, 滚动轴承, 故障诊断, 特征提取