RCMNAAPE在旋转机械故障诊断中的应用

Application of RCMNAAPE in fault diagnosis of rotating machinery

针对精细复合多尺度排列熵(RCMPE)无法充分提取旋转机械振动信号中的故障信息,从而导致旋转机械故障识别准确率不稳定这一缺陷,提出了一种基于精细复合多尺度归一化幅值感知排列熵(RCMNAAPE)、拉普拉斯分数(LS)和灰狼算法优化支持向量机(GWO-SVM)的旋转机械故障诊断方法。首先,利用幅值感知排列熵替换了RCMPE中的排列熵,提出了RCMNAAPE,并将其用于提取旋转机械振动信号的故障特征生成特征样本;随后,采用了LS从原始的高维故障特征向量中筛选出较少的能够更准确描述故障状态的特征,构造敏感特征样本;最后,将低维的故障特征向量输入由灰狼算法优化的支持向量机中进行了训练和测试,完成了旋转机械样本的故障识别和分类,利用滚动轴承和齿轮箱故障数据集将RCMNAAPE-LS-GWO-SVM与其他故障诊断方法进行了对比分析,并开展了评估。研究结果表明:基于RCMNAAPE-LS-GWO-SVM的故障诊断方法能够有效识别旋转机械的各类故障,其识别准确率高于其他对比的故障诊断方法,其中滚动轴承故障的识别准确率达到99.33%,齿轮箱故障的识别准确率达到98.67%。虽然,该方法的特征提取效率不佳,平均特征提取时间分别为153.02 s和163.98 s,仅优于精细复合多尺度模糊熵(RCMFE),但其综合性能更加优异。

Aiming at the defect that refined composite multiscale permutation entropy(RCMPE) could not fully extract fault information from vibration signals of rotating machinery,which led to unstable fault identification accuracy of rotating machinery,a fault diagnosis method for rotating machinery based on refined composite multiscale normalized amplitude aware permutation entropy(RCMNAAPE),Laplace scores(LS) and grey wolf algorithm optimization support vector machine(GWO-SVM) was proposed.Firstly,the amplitude aware permutation entropy was used to replace the permutation entropy in RCMPE,and the RCMNAAPE was proposed to extract the fault characteristics of the vibration signals of rotating machinery and generate the feature samples.Subsequently,LS was used to select fewer features from the original high-dimensional fault feature vectors that can more accurately describe the fault state,and sensitive feature samples were constructed.Finally,the low-dimensional fault feature vector was input into the support vector machine optimized by grey wolf algorithm for training and testing,and the fault identification and classification of rotating machinery samples were completed.The RCMNAAPE-LS-GWO-SVM and other fault diagnosis methods were compared and evaluated by using rolling bearing and gearbox fault data set.The results show that the RCMNAAPE-LS-GWO-SVM fault diagnosis method can effectively identify various kinds of rotating machinery faults,and its recognition accuracy is higher than other fault diagnosis methods,among which the rolling bearing fault recognition accuracy reaches 99.33%,and the identification accuracy of gearbox fault reaches 98.67%.However,the feature extraction efficiency of this method is not good,and the average feature extraction time is respectively 153.02 s and 163.98 s,which is only better than refined composite multiscale fuzzy entropy(RCMFE),but its comprehensive performance is better.