结合狄利克雷过程和连续隐马尔科夫模型的滚动轴承性能退化评估
|
摘要
针对隐马尔科夫模型(Hidden Markov model,HMM)定义中状态数必须预先设定的不足,提出了一种基于狄利克雷过程(Dirichlet process,DP)和连续隐马尔科夫模型(Continuous hidden Markov model,CHMM)的滚动轴承性能退化评估方法。该方法基于DP扩展混合模型(Dirichlet process mixture model,DPMM)良好的聚类特性和分层狄利克雷过程(Hierarchical Dirichlet process,HDP)的分层共享原理,利用多组状态特征值,获得了轴承的运行状态数,解决了CHMM模型结构设置的问题,实现了滚动轴承运行中的退化状态识别和性能评估。利用美国USFI/UCR智能维护中心轴承全寿命试验进行了应用研究,并与基于Kolmogorov-Smirnov(K-S)检验的滚动轴承性能退化评估进行了对比。结果表明,结合狄利克雷过程和连续隐马尔科夫模型的算法能有效地监测滚动轴承运行中的不同退化状态,为基于状态的设备维修提供了参考。
To overcome the deficiency of the definition of Hidden Markov Model(HMM),in which the state number must be set in advance,the performance degradation evaluation method of rolling bearing based on Dirichlet Process(DP)and Continuous Hidden Markov Model(CHMM)was proposed.By the good clustering characteristics of Dirichelet Process Mixture Model(DPMM)and good hierarchical principle of Hierarchical Dirichlet Process(HDP),the degradation state number of rolling bearing was obtained and the CHMM structure was determined.Degradation state recognition and performance evaluation of rolling bearing operation were realized.The application research of the bearing life test of the USFI/UCR intelligent maintenance center in the United States was carried out,and the rolling bearing performance degradation assessment algorithm based on HDP-CHMM and Kolmogorov-Smirnov(K-S)test were compared.Results show that the different degradation states of rolling bearing based on DP and CHMM could be monitored effectively,and it provides reference for the state-based equipment maintenance.
To overcome the deficiency of the definition of Hidden Markov Model(HMM),in which the state number must be set in advance,the performance degradation evaluation method of rolling bearing based on Dirichlet Process(DP)and Continuous Hidden Markov Model(CHMM)was proposed.By the good clustering characteristics of Dirichelet Process Mixture Model(DPMM)and good hierarchical principle of Hierarchical Dirichlet Process(HDP),the degradation state number of rolling bearing was obtained and the CHMM structure was determined.Degradation state recognition and performance evaluation of rolling bearing operation were realized.The application research of the bearing life test of the USFI/UCR intelligent maintenance center in the United States was carried out,and the rolling bearing performance degradation assessment algorithm based on HDP-CHMM and Kolmogorov-Smirnov(K-S)test were compared.Results show that the different degradation states of rolling bearing based on DP and CHMM could be monitored effectively,and it provides reference for the state-based equipment maintenance.
引文
[1]张键.机械故障诊断技术[M].北京:机械工业出版社,2014:129-134.
[2]关健,王黎钦.考虑载荷大小变化次序的球轴承寿命修正模型[J].吉林大学学报:工学版,2016,46(4):1182-1189.Guan Jian,Wang Li-qing.Modified life model of ball bearing considering load changing sequence[J].Journal of Jilin University(Engineering and Technology Edition),2016,46(4):1182-1189.
[3]Peng Y,Dong M.A prognosis method using agedependent hidden semi-Markov model for equipment health prediction[J].Mechanical System and Signal Processing,2011,25(1):237-252.
[4]曾庆虎.机械传动系统关键零部件故障预测技术研究[D].长沙:国防科学技术大学机电工程与自动化学院,2010:65-72.Zeng Qing-hu.Fault prognosties technologies researeh for key parts and components of mechanical transmission systems[D].Changsha:School of Mechantronic Engineering and Automaion,National University of Defense Tcchnology Changsha,2010:65-72.
[5]李庆,Liang S Y,杨建国.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,2016,50(5):723-729.Li Qing,Liang S Y,Yang Jian-guo.Harmonic wavelet sample entropy and HMM model for bearing fault pattern recognition[J].Journal of Shanghai Jiao Tong University,2016,50(5):723-729.
[6]姜万录,杨凯,董克岩,等.基于CHMM的轴承性能退化程度综合评估方法研究[J].仪器仪表学报,2016,37(9):2014-2021.Jiang Wan-lu,Yang Kai,Dong Ke-yan,et al.Research on comprehensive evaluation method of bearing performance degradation based on CHMM[J].Chinese Journal of Scientific Instrument,2016,37(9):2014-2021.
[7]张西宁,雷威,李兵.主分量分析和隐马尔科夫模型结合的轴承监测诊断方法[J].西安交通大学学报,2017,51(6):1-8.Zhang Xi-ning,Lei Wei,Li Bin.Bearing fault detection and diagnosis method based on principal component analysis and hidden Markov model[J].Journal of Xi′an Jiaotong University,2017,51(6):1-8.
[8]Ferguson J D.Variable duration models for speech[C]∥Symposium on the Application of Hidden Markov Models to Text and Speech,1980:143-179.
[9]梅素玉,王飞,周水庚.狄利克雷过程混合模型、扩展模型及应用[J].科学通报,2012,57(34):3243-3257.Mei Su-yu,Wang Fei,Zhou Shui-geng.Dirichlet processes mixture model,extended model and application[J].Chinese Science Bulletin,2012,57(34):3243-3257.
[10]Case western reserve university bearing data center.12kDrive end bearing fault data[DB/OL].[2017-06-22].http:∥csegroups.case.edu/bearingdatacenter/pages/12k-drive-end-bearing-fault-data.
[11]周建英,王飞跃,曾大军.分层Dirichlet过程及其应用综述[J].自动化学报,2011,37(4):389-407.Zhou Jian-ying,Wang Fei-yue,Zeng Da-jun.Review on hierarchical Dirichlet processes and application[J].Acta Automatica Sinica,2011,37(4):389-407.
[12]Teh Y W,Jordan M I,Beal M J,et al.Hierarchical Dirichlet processes[J].Journal of the American Statistical Association,2006,101(476):1566-1581.
[13]Neal R M.Markov chain sampling methods for dirichlet process mixture models[J].Journal of Computational&Graphical Statistics,2000,9(9):249-265.
[14]Lee J,Qiu H,Yu G,et al.'Bearing Data Set′,IMS,University of Cincinnati.NASA ames prognostics data repository[DB/OL].[2017-06-22].http:∥ti.arc.nasa.gov/project/prognostic-data-repository.
[15]王恒,马海波,黄希,等.基于Kolmogrov-Smirnov检验和LS-SVM的机械设备故障预测[J].中南大学学报:自然科学版,2016,47(6):1924-1929.Wang Heng,Ma Hai-bo,Huang Xi,et al.Research on fault prediction of mechanical equipment based on Kolmogrov-Smirnov and LS-SVM[J].Journal of Central South University(Science and Technology),2016,47(6):1925-1929.
[2]关健,王黎钦.考虑载荷大小变化次序的球轴承寿命修正模型[J].吉林大学学报:工学版,2016,46(4):1182-1189.Guan Jian,Wang Li-qing.Modified life model of ball bearing considering load changing sequence[J].Journal of Jilin University(Engineering and Technology Edition),2016,46(4):1182-1189.
[3]Peng Y,Dong M.A prognosis method using agedependent hidden semi-Markov model for equipment health prediction[J].Mechanical System and Signal Processing,2011,25(1):237-252.
[4]曾庆虎.机械传动系统关键零部件故障预测技术研究[D].长沙:国防科学技术大学机电工程与自动化学院,2010:65-72.Zeng Qing-hu.Fault prognosties technologies researeh for key parts and components of mechanical transmission systems[D].Changsha:School of Mechantronic Engineering and Automaion,National University of Defense Tcchnology Changsha,2010:65-72.
[5]李庆,Liang S Y,杨建国.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,2016,50(5):723-729.Li Qing,Liang S Y,Yang Jian-guo.Harmonic wavelet sample entropy and HMM model for bearing fault pattern recognition[J].Journal of Shanghai Jiao Tong University,2016,50(5):723-729.
[6]姜万录,杨凯,董克岩,等.基于CHMM的轴承性能退化程度综合评估方法研究[J].仪器仪表学报,2016,37(9):2014-2021.Jiang Wan-lu,Yang Kai,Dong Ke-yan,et al.Research on comprehensive evaluation method of bearing performance degradation based on CHMM[J].Chinese Journal of Scientific Instrument,2016,37(9):2014-2021.
[7]张西宁,雷威,李兵.主分量分析和隐马尔科夫模型结合的轴承监测诊断方法[J].西安交通大学学报,2017,51(6):1-8.Zhang Xi-ning,Lei Wei,Li Bin.Bearing fault detection and diagnosis method based on principal component analysis and hidden Markov model[J].Journal of Xi′an Jiaotong University,2017,51(6):1-8.
[8]Ferguson J D.Variable duration models for speech[C]∥Symposium on the Application of Hidden Markov Models to Text and Speech,1980:143-179.
[9]梅素玉,王飞,周水庚.狄利克雷过程混合模型、扩展模型及应用[J].科学通报,2012,57(34):3243-3257.Mei Su-yu,Wang Fei,Zhou Shui-geng.Dirichlet processes mixture model,extended model and application[J].Chinese Science Bulletin,2012,57(34):3243-3257.
[10]Case western reserve university bearing data center.12kDrive end bearing fault data[DB/OL].[2017-06-22].http:∥csegroups.case.edu/bearingdatacenter/pages/12k-drive-end-bearing-fault-data.
[11]周建英,王飞跃,曾大军.分层Dirichlet过程及其应用综述[J].自动化学报,2011,37(4):389-407.Zhou Jian-ying,Wang Fei-yue,Zeng Da-jun.Review on hierarchical Dirichlet processes and application[J].Acta Automatica Sinica,2011,37(4):389-407.
[12]Teh Y W,Jordan M I,Beal M J,et al.Hierarchical Dirichlet processes[J].Journal of the American Statistical Association,2006,101(476):1566-1581.
[13]Neal R M.Markov chain sampling methods for dirichlet process mixture models[J].Journal of Computational&Graphical Statistics,2000,9(9):249-265.
[14]Lee J,Qiu H,Yu G,et al.'Bearing Data Set′,IMS,University of Cincinnati.NASA ames prognostics data repository[DB/OL].[2017-06-22].http:∥ti.arc.nasa.gov/project/prognostic-data-repository.
[15]王恒,马海波,黄希,等.基于Kolmogrov-Smirnov检验和LS-SVM的机械设备故障预测[J].中南大学学报:自然科学版,2016,47(6):1924-1929.Wang Heng,Ma Hai-bo,Huang Xi,et al.Research on fault prediction of mechanical equipment based on Kolmogrov-Smirnov and LS-SVM[J].Journal of Central South University(Science and Technology),2016,47(6):1925-1929.