基于精细复合多尺度熵和自编码的滚动轴承故障诊断方法  

Fault Diagnosis Method of Rolling Bearings based on Refined Composite Multiscale Entropy and Autoencoder

作　　者：郑近德 潘海洋[1] 包家汉[1,2] 刘庆运 丁克勤[3] 欧淑彬 ZHENG Jinde;PAN Haiyang;BAO Jiahan;LIU Qingyun;DING Keqin;OU Shubin(Engineering Research Center of Hydraulic Vibration and Control, Ministry of Education, Maanshan 243032, Anhui China;School of Mechanical Engineering, Anhui University of Technology, Maanshan 243032, Anhui China;China Special Equipment Inspection and Research Institute, Beijing 100029, China)

机构地区：[1]液压振动与控制教育部工程研究中心,安徽马鞍山243032 [2]安徽工业大学机械工程学院,安徽马鞍山243032 [3]中国特种设备检测研究院,北京100029

出　　处：《噪声与振动控制》2019年第2期175-180,193,共7页Noise and Vibration Control

基　　金：国家自然科学基金资助项目(51505002);国家重点研发计划资助项目(2017YFC0805100);安徽省高校自然科学研究重大资助项目(KJ2018ZD005);安徽省教育厅自然科学基金重点资助项目(KJ2015A134)

摘　　要：多尺度熵是一种有效衡量机械振动信号复杂度的非线性动力学方法。针对其存在的不足,引入精细复合多尺度熵(Refined composite multiscale entropy, RCMSE),在此基础上,结合自编码降维和遗传优化支持向量机,提出一种滚动轴承故障智能诊断新方法。首先,利用RCMSE提取滚动轴承振动信号多尺度复杂度特征,构建初始特征向量矩阵;其次,采用自编码对初始高维特征数据降维,得到低维流形特征;然后,将低维特征向量输入到基于遗传优化支持向量机的多故障模式分类器中进行训练、识别与诊断。最后,将所提方法应用于实验数据分析,并与多尺度熵方法进行对比,结果表明,该方法不仅能够有效诊断滚动轴承的工作状态和故障类型,而且识别率高于所对比方法。Multi-scale entropy (MSE) is an effective nonlinear dynamics method for complexity measurement of mechanical vibration signals. Aiming at the insufficiency of MSE for shorter time series analysis, the refined composite multi-scale entropy (RCMSE) is introduced. By combining autoencoder for dimension reduction with genetic algorithm optimized support vector machine (GA-SVM), a new intelligent fault diagnosis method for rolling bearings is proposed. Firstly, the RCMSE is used to extract the multi-scale complexity characteristics of vibration signal and construct the initial fault feature matrix. Secondly, the autoencoder is used to reduce the dimension of the initial high-dimensional fault feature data to obtain the low-dimensional manifold features and realize the data visualization. Then, the low-dimensional features are input to the GA-SVM based multi-fault classifier for training, identifying and diagnosis. Finally, the proposed method is applied to the experimental data analysis and compared with the MSE based fault diagnosis method. The results show that the proposed method can effectively diagnose the working state and fault location of rolling bearings with a higher recognition rate than the MSE based fault diagnosis method.

关 键 词：故障诊断 多尺度熵 精细复合多尺度熵 特征降维 滚动轴承 

分 类 号：TH165.3[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]