谐波小波样本熵与HMM模型的轴承故障模式识别

上海交通大学学报（自然版） ›› 2016, Vol. 50 ›› Issue (05): 723-729.

谐波小波样本熵与HMM模型的轴承故障模式识别

李庆1,LIANG Steven Y1,2,杨建国1

（1.东华大学机械工程学院，上海 201620; 2. George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 303320560, USA)

收稿日期:2015-07-01 出版日期:2016-05-28 发布日期:2016-05-28
基金资助:
国家自然科学基金资助项目（51175077），上海市自然科学基金资助项目（14ZR1418500）

Bearing Fault Pattern Recognition Using Harmonic Wavelet Sample Entropy and Hidden Markov Model

LI Qing1,LIANG Steven Y1,2,YANG Jianguo1

（1. College of Mechanical Engineering, Donghua University, Shanghai, 201620, China; 2. George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 303320560, USA)

Received:2015-07-01 Online:2016-05-28 Published:2016-05-28

摘要/Abstract

摘要： 摘要：根据谐波小波分解非平稳振动信号优良特性与隐马尔科夫（HMM）模型的时序模式分类能力，提出了一种基于谐波小波样本熵与HMM模型结合的轴承故障模式识别方法.该方法首先利用谐波小波对轴承各个状态故障信号进行分解，进而由谐波小波三维时频网格图的频率层数特征计算合理的样本熵维数和阈值，依次提取轴承振动信号各层的样本熵构成特征向量序列；然后将序列前120组输入HMM模型中进行训练得到对应故障模型，剩余80组进行测试与识别，通过对比对数似然估计概率输出值确定轴承故障类型.实验通过与BP和RBF神经网络模型进行不同训练组数的正确识别率对比，验证了该组合方法具有识别准确率高，稳定性强的优点.

关键词: 谐波小波, 样本熵, HMM模型, 滚动轴承, 模式识别

Abstract: Abstract： According to the excellent characteristics of the nonstationary vibration signals decompose with harmonic wavelet and the strong temporal pattern classification ability of hidden Markov model(HMM), a bearing fault pattern recognition based on harmonic wavelet sample entropy and HMM was proposed. First, by applying the harmonic wavelet to decompose each bearing fault signal, the frequency layer characteristics of harmonic wavelet threedimensional time frequency trellis was used to estimate the reasonable sample entropy dimension and threshold value, and the feature vector sequence was constructed by extracting sample entropy of rolling bearing each layer vibration signal. Then, the feature vector sequence of former 120 groups were input into HMM to be trained, so as to acquire various models corresponding to different faults，and the remaining 80 groups are tested and identified. Finally, the bearing fault types were identified by comparing the logarithmic likelihood probability value. The actual identify results and comparison with the BP, RBF neural network model prove that this coupling method has better identification accuracy and stability.

Key words: Key words： harmonic wavelet, sample entropy, hidden Markov model(HMM), rolling bearing, pattern recognition

中图分类号:

李庆1,LIANG Steven Y1,2,杨建国1. 谐波小波样本熵与HMM模型的轴承故障模式识别[J]. 上海交通大学学报（自然版）, 2016, 50(05): 723-729.

LI Qing1,LIANG Steven Y1,2,YANG Jianguo1. Bearing Fault Pattern Recognition Using Harmonic Wavelet Sample Entropy and Hidden Markov Model[J]. Journal of Shanghai Jiaotong University, 2016, 50(05): 723-729.

参考文献

［1］INOUE T, SUEOKA A, KANEMOTO H, et al. Detection of minute signs of a small fault in a periodic or a quasiperiodic signal by the harmonic wavelet transform［J］. Mechanical Systems and Signal Processing, 2007, 21(5): 20412055. ［2］RICHMAN J S, LAKE D E, MOORMAN J R. Sample entropy［J］. Methods in Enzymology, 2004,384:172184. ［3］LSCHER P, WWIBEL R, BURGHARDT D. Integrating ontological modelling and Bayesian inference for pattern classification in topographic vector data［J］. Computers, Environment and Urban Systems, 2009, 33(5): 363374. ［4］ZHANG Xuyao, LIU Chenglin. Evaluation of weighted Fisher criteria for large category dimensionality reduction in application to Chinese handwriting recognition［J］. Pattern Recognition, 2013, 46(9): 25992611. ［5］MELIN P, CASTILLO O. A review on the applications of type2 fuzzy logic in classification and pattern recognition［J］. Expert Systems with Applications, 2013, 40(13): 54135423. ［6］SEVER A. Neural network algorithm to pattern recognition in inverse problems［J］. Applied Mathematics and Computation, 2013, 221(15): 484490. ［7］王国锋,李玉波,秦旭达，等. 基于TVARHMM的滚动轴承故障诊断［J］. 天津大学学报, 2010, 43(2):168173. WANG Guofeng, LI Yubo, QIN Xuda, et al. Fault diagnosis of rolling bearing based on TVAR and HMM［J］. Journal of Tianjin University, 2010, 43(2):168173. ［8］DONG Ming，HE D，BANERJEE P，et al. Equipment health diagnosis and prognosis using hidden semiMarkov models［J］. International Journal of Advanced Manufacturing Technology，2006，30(7)：738749. ［9］TAI A H，CHING W K，CHAN L Y. Detection of machine failure：Hidden Markov model approach［J］. Computers & Industrial Engineering, 2009, 57(2):608619. ［10］WANG Jiyan, ZHAO Yucheng, YAO Banghua, et al. Filtering detecting signal of rockbolt with harmonic wavelet［J］. Mining Science and Technology, 2010, 20(3): 411414. ［11］SONG Wanqing, ZHANG Jing. Tool state detection by harmonic wavelet and sample entropy［J］. Chinese Journal of Mechanical Engineering, 2011,24(6):10681073. ［12］PINCUS S M. Assessing serial irregularity and its implications for health［J］. Annals of the New York Academy of Sciences, 2002, 954:24567. ［13］XIE Fengyun, WU Bo, HU Youmin, et al. A generalized interval probabilitybased optimization method for training generalized hidden Markov model［J］. Signal Processing, 2014, 94: 319329. ［14］冯长建.HMM动态模式识别理论、方法以及在旋转机械故障诊断中的应用［D］.杭州：浙江大学机械与能源工程学院，2002. ［15］LLOYD S P. Least squares quantization in PCM［J］. IEEE Transactions on Information Theory, 1982, 28(2): 129137.

[1]	舒俊清, 许昱晖, 夏唐斌, 潘尔顺, 奚立峰. 面向多故障模式的多尺度相似性集成寿命预测[J]. 上海交通大学学报, 2022, 56(5): 564-575.
[2]	卓鹏程, 严瑾, 郑美妹, 夏唐斌, 奚立峰. 面向滚动轴承全生命周期故障诊断的GA-OIHF Elman神经网络算法[J]. 上海交通大学学报, 2021, 55(10): 1255-1262.
[3]	吴军a, b，黎国强a，吴超勇a，程一伟c，邓超c. 数据驱动的滚动轴承性能衰退状态监测方法[J]. 上海交通大学学报（自然版）, 2018, 52(5): 538-544.
[4]	赵秀亮，程勇，王丽梅. 利用模式识别技术提高燃烧激励响应信号信噪比[J]. 上海交通大学学报（自然版）, 2015, 49(05): 637-643.
[5]	李艳峰，王新晴，张梅军，朱会杰. 基于奇异值分解和深度信度网络多分类器的滚动轴承故障诊断方法[J]. 上海交通大学学报（自然版）, 2015, 49(05): 681-686.
[6]	田福庆1，罗荣1，李万2，谢勇1. 改进的谐波小波包峭度图及其应用[J]. 上海交通大学学报（自然版）, 2014, 48(1): 39-44.
[7]	王进，张军，胡白帆. 结合最优类别信息离散的细粒度超网络微阵列数据分类[J]. 上海交通大学学报（自然版）, 2013, 47(12): 1856-1862.
[8]	肖敏1, 廖少明1, 周松2. 基于Web的隧道盾构法施工动态监控系统[J]. 上海交通大学学报（自然版）, 2012, 46(01): 105-108.
[9]	廖雯竹1, 潘尔顺2, 王莹2, 奚立峰2. 统计模式识别和自回归滑动平均模型在设备剩余寿命预测中的应用[J]. 上海交通大学学报（自然版）, 2011, 45(07): 1000-1005.
[10]	杨旭，杨新，熊惠霖. 一种用于支持向量机分类器的组合核优化方法 [J]. 上海交通大学学报（自然版）, 2010, 44(08): 1037-1041.