基于EMD和支持向量机的齿轮箱故障诊断方法研究
|
摘要
机械设备的诊断过程包括诊断信息的获取、故障特征提取和状态识别三部分。其中故障特征提取和状态识别是故障诊断的关键。本文将时频分析的经验模态分解(EMD)和状态识别的支持向量机(SVM)相结合应用于齿轮箱故障诊断当中。
EMD方法是基于信号局部特征时间尺度,能把复杂的信号函数分解为有限个固有模态函数(IMF)之和,而每个IMF表示了原始信号的一个固有振动模态,它们很好地体现了信号的局部特性。此外,由于每一个IMF所包含的频率成分不仅与采样频率有关,而且它还随着信号本身的变化而变化,因此EMD方法是一种自适应的时频局部化分析方法,非常适用于非平稳、非线性的信号处理过程。针对齿轮箱故障振动信号的非平稳特性,本文将EMD方法引入齿轮箱故障特征提取当中,对其基本理论进行研究,实现了用EMD方法提取信号故障特征。并对其端点效应产生失真问题,提出了一种新的解决方法,增强了EMD分解的精确度。
支持向量机具有小样本、较好的泛化能力、全局最优解等特点,在状态识别领域中表现出优良的特性。针对在机械故障诊断中难以获得大量故障典型样本的实际情况,本文在基于支持向量机理论的基础上,开展了对齿轮箱的工作状态和故障类型进行分类和识别研究,并将其研究结果带入实验中,从数据的分析结果表明,EMD和SVM相结合可有效的应用于齿轮箱故障诊断当中。
本文主要工作包括:
1)从齿轮箱零部件常见的失效形式、齿轮系统的振动机理和齿轮箱典型故障振动信号的特征三个方面论述了有关齿轮箱故障诊断的基本知识。
2)从分析EMD方法的基本理论入手,对EMD故障特征提取方法进行研究。其中重点分析了EMD的端点效应带来的影响,提出了处理端点效应的新方法,解决了EMD分解过程中产生失真的现象。
3)从统计学习理论出发,基于支持向量机理论基础开展了状态识别的研究建立起相关的分类器及分类器的算法,验证了其算法的有效性。
4)针对齿轮箱常见的故障,进行故障模拟实验,并应用EMD提取故障信号特征,采用SVM分类器对齿轮箱的工作状态和故障类型进行分类、识别,得到较好的效果。
The process of machinery fault diagnosis includes the acquisition of information and extracting feature and recognizing conditions of which feature extraction and condition identification are the priority. A method of time-frequency analysis, Empirical Mode Decomposition (EMD) and the comparatively recent development of pattern recognition techniques, Support Vector Machines (SVM), are combined and applied to the gearbox fault diagnosis.
EMD is based on the local characteristic time scale of signal and decompose the complicated signal into a number of Intrinsic Mode Functions (IMF). Involving the local characteristic of the signal, each IMF component expresses the original signal of an inherent vibration mode. In addition, the frequency components involved in each IMF not only relates to sampling frequency but also changes with the signal itself, therefore, EMD is a self-adaptive time frequency analysis method that is applicable to non-linear and not-stationary processes perfectly. According to the non-stationary vibration signal characteristics of gearbox, EMD method is introduced into gearbox fault diagnosis in the paper. The basic theory of EMD is studied to extract the signal fault feature. And a new method dealing with the end effect of EMD is proposed to enhance the accuracy of EMD decomposition.
SVM has better generalization and guarantee the local optimal solution is exactly the global optimal solution. SVM can solve the learning problem of a smaller number of samples. Due to the fact that it is difficult to obtain sufficient fault samples in practice, SVM is introduced into gearbox fault diagnosis due to their high accuracy and good generalization for a smaller sample number. Based on the theory of SVM, the work on the gear box status and fault type classification and identification is studied in the paper. And it’s findings into the experiment. The experimental results demonstrate the proposed diagnosis approach in which EMD and SVM are combined is effective.
The outline of the work is as follows:
1. The common parts of gearbox failure, gearbox vibration mechanism and the typical failure characteristics of vibration signals are introduced.
2. The fault feature extraction from the analysis of the EMD basic theory is studied, and the impact of the end effect of EMD is analyzed. A new method dealing with the end effect of EMD is proposed, and it solved the signal distortion during the process of EMD.
3. The status of the classifier to establish relevant and classifier algorithms,and verifies the validity of the algorithm is researched from the statistical learning theory.
4. Simulation experiments of gear box common fault were done to extract the feature of signal with EMD. SVM classifier was used on the gear box of the type of job status and fault classification, and gets a better effect.
EMD方法是基于信号局部特征时间尺度,能把复杂的信号函数分解为有限个固有模态函数(IMF)之和,而每个IMF表示了原始信号的一个固有振动模态,它们很好地体现了信号的局部特性。此外,由于每一个IMF所包含的频率成分不仅与采样频率有关,而且它还随着信号本身的变化而变化,因此EMD方法是一种自适应的时频局部化分析方法,非常适用于非平稳、非线性的信号处理过程。针对齿轮箱故障振动信号的非平稳特性,本文将EMD方法引入齿轮箱故障特征提取当中,对其基本理论进行研究,实现了用EMD方法提取信号故障特征。并对其端点效应产生失真问题,提出了一种新的解决方法,增强了EMD分解的精确度。
支持向量机具有小样本、较好的泛化能力、全局最优解等特点,在状态识别领域中表现出优良的特性。针对在机械故障诊断中难以获得大量故障典型样本的实际情况,本文在基于支持向量机理论的基础上,开展了对齿轮箱的工作状态和故障类型进行分类和识别研究,并将其研究结果带入实验中,从数据的分析结果表明,EMD和SVM相结合可有效的应用于齿轮箱故障诊断当中。
本文主要工作包括:
1)从齿轮箱零部件常见的失效形式、齿轮系统的振动机理和齿轮箱典型故障振动信号的特征三个方面论述了有关齿轮箱故障诊断的基本知识。
2)从分析EMD方法的基本理论入手,对EMD故障特征提取方法进行研究。其中重点分析了EMD的端点效应带来的影响,提出了处理端点效应的新方法,解决了EMD分解过程中产生失真的现象。
3)从统计学习理论出发,基于支持向量机理论基础开展了状态识别的研究建立起相关的分类器及分类器的算法,验证了其算法的有效性。
4)针对齿轮箱常见的故障,进行故障模拟实验,并应用EMD提取故障信号特征,采用SVM分类器对齿轮箱的工作状态和故障类型进行分类、识别,得到较好的效果。
The process of machinery fault diagnosis includes the acquisition of information and extracting feature and recognizing conditions of which feature extraction and condition identification are the priority. A method of time-frequency analysis, Empirical Mode Decomposition (EMD) and the comparatively recent development of pattern recognition techniques, Support Vector Machines (SVM), are combined and applied to the gearbox fault diagnosis.
EMD is based on the local characteristic time scale of signal and decompose the complicated signal into a number of Intrinsic Mode Functions (IMF). Involving the local characteristic of the signal, each IMF component expresses the original signal of an inherent vibration mode. In addition, the frequency components involved in each IMF not only relates to sampling frequency but also changes with the signal itself, therefore, EMD is a self-adaptive time frequency analysis method that is applicable to non-linear and not-stationary processes perfectly. According to the non-stationary vibration signal characteristics of gearbox, EMD method is introduced into gearbox fault diagnosis in the paper. The basic theory of EMD is studied to extract the signal fault feature. And a new method dealing with the end effect of EMD is proposed to enhance the accuracy of EMD decomposition.
SVM has better generalization and guarantee the local optimal solution is exactly the global optimal solution. SVM can solve the learning problem of a smaller number of samples. Due to the fact that it is difficult to obtain sufficient fault samples in practice, SVM is introduced into gearbox fault diagnosis due to their high accuracy and good generalization for a smaller sample number. Based on the theory of SVM, the work on the gear box status and fault type classification and identification is studied in the paper. And it’s findings into the experiment. The experimental results demonstrate the proposed diagnosis approach in which EMD and SVM are combined is effective.
The outline of the work is as follows:
1. The common parts of gearbox failure, gearbox vibration mechanism and the typical failure characteristics of vibration signals are introduced.
2. The fault feature extraction from the analysis of the EMD basic theory is studied, and the impact of the end effect of EMD is analyzed. A new method dealing with the end effect of EMD is proposed, and it solved the signal distortion during the process of EMD.
3. The status of the classifier to establish relevant and classifier algorithms,and verifies the validity of the algorithm is researched from the statistical learning theory.
4. Simulation experiments of gear box common fault were done to extract the feature of signal with EMD. SVM classifier was used on the gear box of the type of job status and fault classification, and gets a better effect.
引文
[1]丁康,李巍华,朱小勇.齿轮及齿轮箱故障诊断实用技术[M].北京:机械工业出版社,2005.5: 1,3,10-11,91-108.
[2]许昕.齿轮箱故障诊断在安全生产中的应用[D].太原:中北大学,2007.
[3]杨宇.基于EMD和支持向量机的旋转机械故障诊断方法研究[D].长沙:湖南大学,2005.
[4]虞和济.机械设备故障诊断的人工神经网络识别法[J].机械强度,1995,17(2):48-54.
[5] Norden E. Huang, Zheng Shen, Steven R. Long, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceeding of Royal Society London, 1998, 454 (A): 903-995.
[6] Huang Norden E., Shen Zheng, Long Steven R., et al. A New View of Nonlinear Water Waves– The Hilbert Spectrum.[J]. Ann. Rev. Fluid Mech., 1999, 31: 417-457.
[7]于德介,程军圣,杨宇.机械故障诊断的Hilbert-Huang变化方法[M].北京:科学出版社,2006.5: 24-34,154-166.
[8]于德介,程军圣. EMD方法在齿轮故障诊断中的应用[J].湖南大学学报,2002,29(6):48-51
[9]李国华,张永忠.机械故障诊断[M].北京:化学工业出版社,1999.1
[10]顾小军.面向旋转机械的支持向量机方法及智能故障诊断系统研究[D]杭州:浙江大学,2006.
[11]杨皓.分形理论应用于齿轮箱滚动轴承故障诊断的研究[D].长沙:长沙理工大学,2007.
[12]成琼.基于小波分析的齿轮故障诊断研究[D].长沙:湖南大学,2001.
[13]王佃武.基于小波分析的齿轮箱齿轮点蚀故障诊断[D].兰州:兰州理工大学,2006.
[14]时建峰基于时、频域-小波分析和神经网络方法的齿轮箱故障诊断研究.[D].太原:太原理工大学,2008.
[15]于德介,程军圣. EMD方法在齿轮故障诊断中的应用[J].湖南大学学报,2002,29(6):48-51.
[16] Chen Q. H., Huang N. E., Long Steven R., et al. A B-spline approach for empirical mode decompositions[J]. Advances in Computational Mathematics, 2006 , 24: 171-195.
[17]程军圣.基于Hilbert-Huang变换的旋转机械故障诊断方法研究[D].长沙:湖南大学,2005.
[18]黄大吉,赵进平,苏纪兰.希尔伯特-黄变换的端点延拓[J].海洋学报, 2003.1,25(1):1-11.
[19] G. Rilling, P. Flandrin, P. Goncalyes. On empirical mode decomposition and its algorithms.[C]. //IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing(NSIP2003), Grado(I), June, 2003: 8-11.
[20]舒忠平,杨智春.抑制经验模分解边缘效应的极值点对称延拓法[J].西北工业大学学报, 2006.10,24(5):639-643.
[21]刘慧婷,张旻,程家兴.基于多项式拟合算法的EMD端点问题的处理[J].计算机工程与应用,2004,16:84-87.
[22]瞿伟廉,程磊.应用径向基函数神经网络处理EMD方法中的边界问题[J].华中科技大学学报,2006.12,23(4):1-5.
[23]杨建文,贾民平.希尔伯特-黄谱的端点效应分析及处理方法研究[J].振动工程学报,2006.6,19(2):283-288.
[24]胡劲松,杨世锡. EMD方法基于AR模型预测的数据延拓与应用[J].振动、测试与诊断,2007.6,27(2):116-121.
[25]苏玉香,刘志刚,李科亮等.一种改善EMD端点效应的新方法及其在谐波分析中的应用[J].电工电能新技术,2008.4,27(2):33-37.
[26]胡爱军,安连锁,唐贵基. HILBERT-HUANG变换端点效应处理新方法[J].机械工程学报,2008.4,44(4):154-158.
[27]任达千,吴昭同,严拱标. EMD端点效应的评价指标及抑制端点效应的窗函数法[J].制造业自动化,2007.1,29(1):21-24.
[28] Huang, N. E., M. L. Wu, S. R. Long, S. S. Shen, W. D. Qu, P. Gloersen, and K. L. Fan. A confidence limit for the Empirical Mode Decomposition and Hilbert Spectral Analysis[J]. Proc. Roy. Soc. London, 2003,459A:2317-2345.
[29] Wu, Z., and N. E. Huang . A study of the characteristics of white noise using the empirical mode decomposition method[J], Proc. Roy. Soc. London, 2004, 460A: 1597-1611.
[30]徐宝杰,张建民,徐小力等.抑制EMD端点效应方法的研究[J].北京理工大学学报,2006.3,26(3):196-200.
[31] Flandrin, P., G. Rilling, and P. Goncalves. Empirical mode decomposition as a filterbank[J]. IEEE Signal Proc Lett.,2004,11:112-114.
[32]方瑞明.支持向量机理论及其应用分析[M].北京:中国电力出版社,2007.10:1,9.
[33]魏于凡.支持向量机在智能故障诊断中的应用研究[D].保定:华北电力大学,2007.
[34]张周锁,李凌均,何正嘉.基于支持向量机的机械故障诊断方法研究[J].西安交通大学学报,2002.12,36(12):1303-1306.
[35]黄景涛.支持向量机算法参数选择及其在电站锅炉系统中的应用[D].杭州:浙江大学,2005.
[36]郭丽娟,孙世宇,段修生.支持向量机及核函数研究[J].科学技术与工程,2008.1,8(2):487-490.
[37] Keerthi S S. Efficient tuning of SVM hyperparameters using radius/margin bound and iterative algorithms[J]. IEEE Trans. On Neural Networks , 2002, 13 (5):1225 -1229.
[38] Steve R. Gunn. Support Vector Machines for Classification and Regression[R]. Technical Report , Southampton: University of Southampton,1998,1-28.
[39]陈丹.多类支持向量机算法的研究[J].东莞理工学院学报,2007.10,14(5):65-69.
[40]段江涛,李凌均,张周锁.基于支持向量机的机械系统多故障分类方法[J].农业机械,2007.4,35(4):144-147.
[41] Chapelle O,Vapnik V, Bousquet O,et al. Choosing multiple parameters for support vector machines[J]. Machine Learning , 2002 ,46 :131-159.
[42] Keerthi S S. Efficient tuning of SVM hyperparameters using radius/margin bound and iterative algorithms[J]. IEEE Trans. On Neural Networks , 2002, 13 (5):1225 -1229.
[43]郑永涛,刘玉树.支持向量机解决多分类问题研究[J].计算机工程与应用,2005,23:190-192.
[44]程圣军,于德介,杨宇.基于EMD的齿轮故障识别研究[J].电子与信息学报,2004.5,26(5):825-829.
[45]李辉,郑海起,唐力伟.应用Hilbert-Huang变换的齿轮破损障诊断研究[J].振动、测试与诊断,2005.9,25(3):200-205.
[46]程圣军,于德介,杨宇.一种基于SVM和EMD的齿轮故障诊断方法[J].机械工程学报,2005.1,41(1):140-144.
[2]许昕.齿轮箱故障诊断在安全生产中的应用[D].太原:中北大学,2007.
[3]杨宇.基于EMD和支持向量机的旋转机械故障诊断方法研究[D].长沙:湖南大学,2005.
[4]虞和济.机械设备故障诊断的人工神经网络识别法[J].机械强度,1995,17(2):48-54.
[5] Norden E. Huang, Zheng Shen, Steven R. Long, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceeding of Royal Society London, 1998, 454 (A): 903-995.
[6] Huang Norden E., Shen Zheng, Long Steven R., et al. A New View of Nonlinear Water Waves– The Hilbert Spectrum.[J]. Ann. Rev. Fluid Mech., 1999, 31: 417-457.
[7]于德介,程军圣,杨宇.机械故障诊断的Hilbert-Huang变化方法[M].北京:科学出版社,2006.5: 24-34,154-166.
[8]于德介,程军圣. EMD方法在齿轮故障诊断中的应用[J].湖南大学学报,2002,29(6):48-51
[9]李国华,张永忠.机械故障诊断[M].北京:化学工业出版社,1999.1
[10]顾小军.面向旋转机械的支持向量机方法及智能故障诊断系统研究[D]杭州:浙江大学,2006.
[11]杨皓.分形理论应用于齿轮箱滚动轴承故障诊断的研究[D].长沙:长沙理工大学,2007.
[12]成琼.基于小波分析的齿轮故障诊断研究[D].长沙:湖南大学,2001.
[13]王佃武.基于小波分析的齿轮箱齿轮点蚀故障诊断[D].兰州:兰州理工大学,2006.
[14]时建峰基于时、频域-小波分析和神经网络方法的齿轮箱故障诊断研究.[D].太原:太原理工大学,2008.
[15]于德介,程军圣. EMD方法在齿轮故障诊断中的应用[J].湖南大学学报,2002,29(6):48-51.
[16] Chen Q. H., Huang N. E., Long Steven R., et al. A B-spline approach for empirical mode decompositions[J]. Advances in Computational Mathematics, 2006 , 24: 171-195.
[17]程军圣.基于Hilbert-Huang变换的旋转机械故障诊断方法研究[D].长沙:湖南大学,2005.
[18]黄大吉,赵进平,苏纪兰.希尔伯特-黄变换的端点延拓[J].海洋学报, 2003.1,25(1):1-11.
[19] G. Rilling, P. Flandrin, P. Goncalyes. On empirical mode decomposition and its algorithms.[C]. //IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing(NSIP2003), Grado(I), June, 2003: 8-11.
[20]舒忠平,杨智春.抑制经验模分解边缘效应的极值点对称延拓法[J].西北工业大学学报, 2006.10,24(5):639-643.
[21]刘慧婷,张旻,程家兴.基于多项式拟合算法的EMD端点问题的处理[J].计算机工程与应用,2004,16:84-87.
[22]瞿伟廉,程磊.应用径向基函数神经网络处理EMD方法中的边界问题[J].华中科技大学学报,2006.12,23(4):1-5.
[23]杨建文,贾民平.希尔伯特-黄谱的端点效应分析及处理方法研究[J].振动工程学报,2006.6,19(2):283-288.
[24]胡劲松,杨世锡. EMD方法基于AR模型预测的数据延拓与应用[J].振动、测试与诊断,2007.6,27(2):116-121.
[25]苏玉香,刘志刚,李科亮等.一种改善EMD端点效应的新方法及其在谐波分析中的应用[J].电工电能新技术,2008.4,27(2):33-37.
[26]胡爱军,安连锁,唐贵基. HILBERT-HUANG变换端点效应处理新方法[J].机械工程学报,2008.4,44(4):154-158.
[27]任达千,吴昭同,严拱标. EMD端点效应的评价指标及抑制端点效应的窗函数法[J].制造业自动化,2007.1,29(1):21-24.
[28] Huang, N. E., M. L. Wu, S. R. Long, S. S. Shen, W. D. Qu, P. Gloersen, and K. L. Fan. A confidence limit for the Empirical Mode Decomposition and Hilbert Spectral Analysis[J]. Proc. Roy. Soc. London, 2003,459A:2317-2345.
[29] Wu, Z., and N. E. Huang . A study of the characteristics of white noise using the empirical mode decomposition method[J], Proc. Roy. Soc. London, 2004, 460A: 1597-1611.
[30]徐宝杰,张建民,徐小力等.抑制EMD端点效应方法的研究[J].北京理工大学学报,2006.3,26(3):196-200.
[31] Flandrin, P., G. Rilling, and P. Goncalves. Empirical mode decomposition as a filterbank[J]. IEEE Signal Proc Lett.,2004,11:112-114.
[32]方瑞明.支持向量机理论及其应用分析[M].北京:中国电力出版社,2007.10:1,9.
[33]魏于凡.支持向量机在智能故障诊断中的应用研究[D].保定:华北电力大学,2007.
[34]张周锁,李凌均,何正嘉.基于支持向量机的机械故障诊断方法研究[J].西安交通大学学报,2002.12,36(12):1303-1306.
[35]黄景涛.支持向量机算法参数选择及其在电站锅炉系统中的应用[D].杭州:浙江大学,2005.
[36]郭丽娟,孙世宇,段修生.支持向量机及核函数研究[J].科学技术与工程,2008.1,8(2):487-490.
[37] Keerthi S S. Efficient tuning of SVM hyperparameters using radius/margin bound and iterative algorithms[J]. IEEE Trans. On Neural Networks , 2002, 13 (5):1225 -1229.
[38] Steve R. Gunn. Support Vector Machines for Classification and Regression[R]. Technical Report , Southampton: University of Southampton,1998,1-28.
[39]陈丹.多类支持向量机算法的研究[J].东莞理工学院学报,2007.10,14(5):65-69.
[40]段江涛,李凌均,张周锁.基于支持向量机的机械系统多故障分类方法[J].农业机械,2007.4,35(4):144-147.
[41] Chapelle O,Vapnik V, Bousquet O,et al. Choosing multiple parameters for support vector machines[J]. Machine Learning , 2002 ,46 :131-159.
[42] Keerthi S S. Efficient tuning of SVM hyperparameters using radius/margin bound and iterative algorithms[J]. IEEE Trans. On Neural Networks , 2002, 13 (5):1225 -1229.
[43]郑永涛,刘玉树.支持向量机解决多分类问题研究[J].计算机工程与应用,2005,23:190-192.
[44]程圣军,于德介,杨宇.基于EMD的齿轮故障识别研究[J].电子与信息学报,2004.5,26(5):825-829.
[45]李辉,郑海起,唐力伟.应用Hilbert-Huang变换的齿轮破损障诊断研究[J].振动、测试与诊断,2005.9,25(3):200-205.
[46]程圣军,于德介,杨宇.一种基于SVM和EMD的齿轮故障诊断方法[J].机械工程学报,2005.1,41(1):140-144.