基于VMD-Hilbert边际谱能量熵和SVM的高压断路器机械故障诊断
来源:用户上传
作者:
摘 要:针对高压断路器分、合闸动作过程中产生的振动信号持续时间短暂及强烈的非线性非平稳性,导致的特征提取困难问题,提出一种变分模态分解(VMD)-希尔伯特(Hilbert)边际谱能量熵,及支持向量机(SVM)的高压断路器振动信号组合特征提取和机械故障诊断方法。采用VMD对高压断路器振动信号进行分解,得到一系列反映振动信号局部特性的本征模态函数(IMF);对IMF进行Hilbert变换,并求取对高压断路器机械状态变化敏感的Hilbert边际谱能量熵作为特征向量;将特征向量输入到SVM分类器,实现高压断路器机械故障的智能诊断。试验结果表明:该方法能够准确识别高压断路器的常见机械故障类型,具有一定的工程应用价值。
关键词:高压断路器;变分模态分解;希尔伯特边际谱;能量熵;支持向量机;机械故障识别
DOI:10.15938/j.emc.2020.03.002
中图分类号:TM 561文献标志码:A文章編号:1007-449X(2020)03-0011-09
Abstract:In this paper, a feature extraction method and fault diagnosis for high voltage circuit breakers (HVCBs) is presented and discussed. The vibration signals are nonlinear and timevarying since the complicated structure and extremely fast operation of HVCBs, which makes the extraction and selection of sensitive features for fault diagnosis difficult. Therefore, it is of vital importance to explore a new vibration feature extraction algorithm to improve the accuracy of fault diagnosis for HVCBs. A combination feature extraction method based on variational mode decomposition (VMD) and Hilbert marginal spectrum energy entropy, and support vector machine (SVM) for the diagnosis of HVCB’s mechanical condition is presented and clearly discussed. Vibration signals were decomposed into several intrinsic mode functions (IMFs) by using VMD. Marginal spectral energy entropies of IMFs (which vary with different fault types of HVCB) were obtained and served as feature vectors for the SVM classifier for the diagnosis of HVCB. Experimental results indicate that the proposed method can accurately identify the common mechanical faults of HVCB and has potential of practical application.
Keywords:high voltage circuit breakers; variational mode decomposition; Hilbert marginal spectrum; energy entropy; support vector machine; mechanical fault detection
0 引 言
高压断路器的可靠性对于保障电力系统的安全稳定运行具有重要的作用。运行实践表明,机械故障是导致高压断路器故障的主要原因。近年来,对高压断路器机械故障诊断的研究越来越多,一些研究成果也已用于实际工程,其中,基于振动信号的高压断路器机械故障诊断技术越来越受到人们的关注[1-3]。
高压断路器分、合闸动作时产生的振动信号蕴含着丰富、重要的高压断路器状态信息[4-6]。由于高压断路器动作时间极短(常常是几十毫秒)、各运动件之间强烈碰撞冲击等特点质,使得其振动信号具有时域时间短、频域分布宽、强烈的非线性非平稳性。所以,一方面,对传感器的性能提出了更高的要求:传感器必须具有足够高的采样精确度,且频响范围及量程应足够大;另一方面,对振动信号的处理也提出了更高的要求,传统的信号处理方法不能有效提取高压断路器这种具有强冲击时变特性振动信号的关键信息。
针对高压断路器振动信号的特殊性,时频分析方法无疑是较适合的,因此,越来越多的时频分析方法被用于分析高压断路器的振动信号。如小波变换[7-9]、经验模态分解[10-12](empirical mode decomposition,EMD)。实际上,小波变换的本质还是一种傅里叶变换,存在信号能量泄漏、基函数选择等问题,且不具备自适应性。EMD是一种可以根据信号自身特点进行自适应多分辨率分解的信号分析方法,但其在分解过程中容易产生模态混叠、本征模态函数(intrinsic mode function,IMF)筛分停止条件和端点效应等问题[13-14]。而变分模态分解[15-17](variational mode decomposition,VMD)通过寻找约束变分模型最优解实现信号的分解,各IMF分量中心频率和带宽不断交替迭代更新,实现信号频带的自适应分解。VMD方法克服了EMD方法的诸多缺陷(如模态混叠等),大大提高信号分解的准确性。振动信号经VMD处理得到一系列反映振动信号局部特性的本征模态函数(IMF);IMF通过希尔伯特(Hilbert)变换可更有效、更真实地获得振动信号中所含的重要信息,即Hilbert谱(Hilbert谱可精确地描述信号幅值在整个频段上随时间和频率的变化规律)。 当高压断路器状态发生变化时,振动信号的频谱(本文采用分辨率比Fourier谱更高的Hilbert边际谱)和能量分布随之发生改变,又由于熵在物理意义上是混乱程度的度量,在能量理论中,熵与信号或随机事件的不确定程度有关,因而,为了能够更好地反映高压断路器状态的变化,引入Hilbert边际谱能量熵(基于能量熵的振动信号分析方法可参阅文献[18-21]),即进一步对Hilbert边际谱求取能量熵。最后,将这些携带着高压断路器振动信号重要信息的IMF Hilbert边际谱能量熵作为特征向量,输入到支持向量机[2,22-23](support vector machine,SVM)(SVM是建立在VC维理论[24]和结构风险最小化[25]基础上的一种有监督学习方法,具有较强的分类能力,在解决小样本、高维非线性决策问题时具有无可比拟的优势。综合考虑实际工程应用中高压断路器故障振动信号样本較少及SVM能够解决小样本、非线性和高维模式识别等优点,选择SVM作为断路器故障诊断方法)中,实现高压断路器的机械状态识别与故障诊断。
本文首先采用VMD对高压断路器振动信号进行分解,得到反映振动信号局部特性的本征模态函数(IMF);然后对各IMF分量进行Hilbert变换并求取Hilbert边际谱的能量值,得到反映振动信号的动态特征向量——Hilbert边际谱能量熵;最后将特征向量输入经过优化的SVM网络中进行机械故障诊断。试验结果表明,本文所提方法能够较准确地判断高压断路器常见机械故障类型,具有一定的工程使用价值。
由此可知,对于同一型号的高压断路器,在样本较少的情况下,SVM方法对本文设定的3类机械故障及正常状态的分类结果较佳。对于不同型号的高压断路器,由于高压断路器结构不同,产生的振动信号亦不同。因此,对于不同型号的高压断路器,若采用本文方法,需进一步验证。但无论是何种类型的高压断路器,其故障机理是一致的:当高压断路器状态发生变化时,振动信号的频谱和能量分布将随之发生改变。因此,本文通过分析振动信号频谱和能量分布的变化规律对高压断路器进行故障诊断的思路和方法是具有普遍意义的。
6 结 论
本文针对高压断路器振动信号的时变性、非线性和非平稳性等特点,导致特征提取困难的问题,提出了一种变分模态分解(VMD)-希尔伯特(Hilbert)边际谱能量熵的组合特征提取及支持向量机(SVM)的高压断路器机械故障诊断方法,得到以下结论:
1)将表征高压断路器振动信号局部特性的本征模态函数(IMF)和反映高压断路器状态特征的能量熵相结合,提取出的振动信号特征值能够准确反映高压断路器的状态变化。
2)VMD处理后得到的各IMF分量Hilbert边际谱比EMD处理得到的IMF分量Fourier谱、Hilbert边际谱频率聚集性更好、分辨率更高,故障特征频段更容易区分。
3)各IMF分量的Hilbert边际谱能量熵具有明显的差异,冗余度较小,能够从不同尺度、更加精确地反映高压断路器的状态信息。
4)基于VMDHilbert边际谱能量熵的振动信号组合特征提取和SVM的高压断路器机械故障诊断方法准确识别了高压断路器3种常见机械故障及正常状态。
参 考 文 献:
[1] LANDRY M, LEONARD F, LANDRY C, et al. An improved vibration analysis algorithm as a diagnostic tool for detecting mechanical anomalies on power circuit breakers[J]. IEEE Transactions on Power Delivery, 2008, 23(4): 1986.
[2] 程序, 关永刚, 张文鹏, 等. 基于因子分析和支持向量机算法的高压断路器机械故障诊断方法[J]. 电工技术学报, 2017, 29(7): 209.
CHENG Xu, GUAN Yonggan, ZHANG Wenpeng, et al. Diagnosis method on the mechanical failure of high voltage circuit breakers based on factor analysis and SVM[J]. Transactions of China Electrotechnical Society, 2017, 29(7): 209.
[3] 赵书涛, 王亚潇, 孙会伟, 等. 基于自适应权重证据理论的断路器故障诊断方法研究[J].中国电机工程学报, 2017, 37(23): 7040.
ZHAO Shutao, WANG Yaxiao, SUN Huiwei, et al. Research of circuit breaker fault recognition method based on adaptive weighted of evidence theory[J]. Proceedings of the CSEE, 2017, 37(23): 7040.
[4] LAI M L, PARK S Y, LIN G C, et al. Mechanical failure detection of circuit breakers[J]. IEEE Transactions on Power Delivery, 1988, 3(4): 1724.
[5] DEMJANENKO V, VALTIN R A, SOUMEKH M, et al. A noninvastve diagnostic instrument for power circuit breakers[J]. IEEE Transactions on Power Delivery, 1992, 7(2): 656. [6] RUNDE M, OTTESEN G E, SKYBERG B,et al. Vibration analysis for diagnostic testing of circuitbreakers[J]. IEEE Transactions on Power Delivery, 1996, 11(4): 1816.
[7] LEE D S S, LITHGOW B J, MORRISON R E. New fault diagnosis of circuit breakers[J]. IEEE Transactions on Power Delivery, 2003, 18(2): 454.
[8] CHARBKAEW N, SUWANASR T, BUNYAGUL T, et al. Vibration signal analysis for condition monitoring of puffertype highvoltage circuit breakers using wavelet transform[J]. IEEJ Transactions on Electrical and Electronic Engineering, 2012, 7(1):13.
[9] LIU Mingliang, WANG Keqi, SUN Laijun, et al. Fault diagnosis method of HV circuit breaker based on wavelets neural network[J]. The Open Automation and Control Systems Journal, 2015, 7:126.
[10] HUANG Jian, HU Xiaoguang, GENG Xin. An intelligent fault diagnosis method of high voltage circuit breaker based on improved EMD energy entropy and multiclass support vector machine[J]. Electric Power Systems Research, 2011, 81: 400.
[11] 张丽萍, 石敦义, 缪希仁. 低压断路器振动特性分析及其故障诊断研究[J]. 电机与控制学报, 2016, 20(10):82.
ZHANGLiping, SHI Dunyi, MIAO Xiren. Research on vibration signal feature analysis and its fault diagnosis[J]. Electric Machines and Control, 2016, 20(10): 82.
[12] 李建鹏, 赵书涛, 夏燕青. 基于双谱和希尔伯特-黄变换的断路器故障诊断方法[J]. 电力自动化设备, 2013, 33(2):115.
LI Jianpeng, ZHAO Shu Tao, XIA Yanqing. Machinery fault diagnosis of high voltage circuit breaker based on empirical mode decomposition[J]. Electric Power Automation Equipment, 2013, 33(2):115.
[13] HUANG N E, SHEN Z, LONG S R, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and nonstationary time series analysis[C]//Proc. Roy. Soc. London Ser. A, 1998, 454: 903-995.
[14] 任宜春, 翁璞. 基于改进HilbertHuang变换的结构损伤识别方法研究[J]. 振动与冲击, 2015, 34(18): 195.
REN Yichun, WENG Pu. Structural damage detection based on improved HilbertHuang transform[J]. Journal of Vibration and Shock, 2015, 34(18): 195.
[15] DRAGOMIRETSKIY K, ZOSSO D. Variational mode decomposition[J]. IEEE Transactions on Signal Processing, 2014, 62(3): 531.
[16] 徐元博, 蔡宗琰, 胡永彪, 等. 強噪声背景下频率加权能量算子和变分模态分解在轴承故障提取中的应用[J]. 振动工程学报, 2018, 31(3): 513.
XU Yuanbo,CAI Zongyan,HU Yongbiao,et al.A frequencyweighted energy operator and variational mode decomposition for bearing fault detection[J].Journal of Vibration Engineering,2018,31(3):513.
[17] 赵岩, 朱均超, 张宝峰, 等. 基于VMD与Hilbert谱的旋转机械碰摩故障诊断方法[J]. 振动、测试与诊断, 2018, 38(2): 381. ZHAO Yan, ZHUYunchao, ZHANG Baofeng, et al. Runimpact fault diagnosis of rotating machinety based on VMD and Hilbert spectrum[J]. Journal of Vibration, Measurement & Diagnosis, 2018, 38(2): 381.
[18] 艾延廷, 费成巍. 转子振动故障的小波能谱熵SVM诊断方法[J]. 航空动力学报, 2011, 26(8):1830.
AI Yanting, FEI Chenwei. Rotor vibration fault diagnosis method based on wavelet energy spectrum entropy and SVM[J]. Journal of Aerospace Power, 2011, 26(8): 1830.
[19] 徐建源, 张彬, 林莘, 等. 能谱熵向量法及粒子群优化的RBF神经网络在高压断路器机械故障诊断中的应用[J]. 高电压技术, 2012, 38(6): 1299.
XU Jianyuan, ZHANG Bin, LIN Xin, et al. Application of energy spectrum entropy vector method and RBF neural networks optimized by the particla swarm in highvoltage circuit breaker mechanical fault diagnosis[J]. High Voltage Engineering, 2012, 38(6): 1299.
[20] LIU Mingliang, WANG Keqi, SUN Laijun, et al. Applying energyequal entropy of wavelet packet to diagnose circuit breaker faults[J]. The Open Electrical & Electronic Engineering Journal, 2014, 8: 445.
[21] 张彬, 徐建源, 陈江波, 等. 基于电力变压器振动信息的绕组形变诊断方法[J]. 高电压技术, 2015, 41(7): 2341.
ZHANG Bin, XU Jianyuan, CHEN Jiangbo, et al. Diagnosis method of winding deformation based on transformer vibration information[J]. High Voltage Engineering, 2015, 41(7): 2341.
[22] VAPNIK V N. Statistical Learning Theory[M]. Hoboken, NJ, USA: Wiley, 1998.
[23] REN Likun, L Weimin, JIANG Shiwei, et al. Fault diagnosis using a joint model based on sparse representation and SVM[J]. IEEE Transactions on Instrumentations and Measurement, 2016, 65(10): 2313.
[24] 刘学艺, 宋春跃, 李平. 基于VapnikChervonenkis泛化界的极限学习机模型复杂性控制[J]. 控制理论与应用, 2014, 31(5): 644.
LIU Xueyi, SONG Chunyue, LI Ping. Model complexity control of extreme learning machine using VapnikChervonenkis generalization bounds[J]. Control Theory & Applications, 2014, 31(5): 644.
[25] ANGUITA D, GHIO A, ONETO L, et al. Insample and outofsample model selection and error estimation for support vector machines[J]. IEEE Transactions on Neural Networks and Learning Systems, 2012, 23(9): 1390.
(編辑:贾志超)
转载注明来源:https://www.xzbu.com/1/view-15174129.htm