基于IRCMNDE和NNCHC的滚动轴承故障诊断

针对多尺度散布熵（MDE）在粗粒化过程中易发生信息丢失、产生虚假信息，难以全面提取轴承故障信息的问题，提出了基于改进的精细复合多尺度归一化散布熵（IRCMNDE）和最近邻凸包分类（NNCHC）的滚动轴承故障诊断方法。引入精细复合多尺度散布熵（RCMDE），将其粗粒化过程中平均值替换为最大值来表示数据段信息，以克服传统粗粒化过程的不足并突出故障特征。通过归一化操作减弱熵值计算时不同参数选择导致的熵值波动幅度，得到IRCMNDE。将IRCMNDE作为故障特征，使用NNCHC分类器对故障特征进行分类。经实验验证，该方法可达到98.98%的故障识别准确率，相比基于MDE（故障识别准确率为95.99%）和RCMDE（故障识别准确率为97.60%）的方法，能够更准确地提取滚动轴承的故障特征信息，提高承故障分类的准确性。 

Fault diagnosis of rolling bearings based on IRCMNDE and NNCHC

In view of the problem that information loss and false information may occur during the coarse-graining process of multi-scale dispersion entropy （MDE），which make it difficult to extract bearing fault information comprehensively，a rolling bearing fault diagnosis method based on improved refined composite multi-scale normalized dispersion entropy （IRCMNDE） and nearest neighbor convex hull classification （NNCHC） was proposed.The refined composite multi-scale dispersion entropy （RCMDE） was introduced，and the average value in the coarse-graining process was replaced by the maximum value to represent the data segments information，which can overcome the shortcomings of the traditional coarse-graining process and highlight the fault characteristics.Through the normalization operation to reduce the influence of the selection of different parameters on the entropy value，IRCMNDE was acquired as feature samples； NNCHC was used to classify the feature samples to realize bearing fault diagnosis.Experimental results showed that the proposed method can achieve 98.98% fault identification accuracy.Compared with the methods based on MDE （fault identification accuracy was 95.99%） and RCMDE （fault identification accuracy was 97.60%），the proposed method can extract the fault feature information of rolling bearings more accurately and improve the accuracy of fault classification.