风电机组传动链振动分析与故障特征提取方法研究

发布时间：2018-03-10 01:23

本文选题：风力发电机组　切入点：传动链　出处：《华北电力大学》2013年博士论文　论文类型：学位论文

【摘要】：振动监测是机械设备状态监测与故障诊断的主要技术之一,在许多工业行业得到广泛应用。我国早期建设的风电场出于降低成本的考虑,一般不配备风电机组振动监测系统。随着设备运行时间的不断延续,风电机组传动链故障率及维修费用高的问题逐渐突出,风电企业开始重视风电机组的运行状态监测及故障诊断,特别是国家能源局于2011年颁布的推荐性国家能源行业标准NB/T31004-2011《风力发电机组振动状态监测导则》,对于风电机组振动监测与故障诊断技术的发展起到了有力的推动作用。风电机组传动链主要由传动轴及支撑轴承、增速齿轮箱等旋转部件组成,是连接风轮和发电机,实现能量转换和传递过程的关键部件。与一般机械设备中的传动链相比,风电机组传动链的载荷状态、运行工况、环境条件和结构布局等方面均比较复杂,导致传动链中齿轮、轴承等主要零部件的故障机理和故障发展模式等存在特殊性,故障率高于其它行业的同类设备,实际使用寿命远远低于设计寿命。因此,针对风电传动链的运行特点及其故障诊断的共性问题,研究新的理论和技术解决方法,提高故障诊断和预测准确性,对于保证风电机组设备的健康运行具有重要意义。风电机组传动链是机、电、液耦合的复杂结构,可能产生故障的部位多,从故障激励源到振动监测点经过不同传递途径的衰减和混合作用,监测到的振动信号往往是不同激励源及传递途径的复杂卷积混合作用,具有背景噪声干扰大、非平稳、非线性的特点,在许多实际场合,用经典的振动信号分析方法难以给出反映故障特征的准确信息,影响了对故障的精确分析诊断,更限制了自动故障诊断和预测技术的应用。因此,根据风电机组传动链中齿轮箱和支撑轴承的特殊性,研究在经典振动信号分析方法的基础上的改进方法,提高故障特征分析与提取功能和效果,是实现准确故障诊断的技术关键。本文在这样的背景下,探讨了对几种经典振动信号分析的改进,并分别用双馈式和直驱式风电机组传动链的实测信号进行验证。主要研究内容和形成的结论如下： (1)倒频谱具有一个“同态滤波”的重要性质,可以实现卷积混合信号的分离。利用倒频谱的同态滤波性质,提出一种提取齿轮箱故障特征的方法,由于监测到的振动信号是激励源与传递过程的结构固有振动的卷积结果,这两种信号成分在倒频域的特性存在明显差别。据此,对齿轮箱振动信号在倒频域内进行带通滤波处理,将激励源成分和结构固有振动成分进行分离,然后对具有低倒频性质的结构共振成分进行频谱重构,用重构频谱反映故障引起的结构固有特性的变化,从而揭示故障发生和发展的趋势。对风电齿轮箱和轴承振动的分析实例表明,该方法提取的特征值具有受运行工况影响小,在各种运行工况下都能够清楚表征故障状态,而且可以很好地反映故障的发生发展趋势。 (2)基于Hilbert变换的窄带包络分析通过选择振动信号中反映结构共振的频带进行窄带滤波,获取反映故障的幅值包络信息。该方法的主要问题是窄带滤波频带不好确定,特别是对于风电传动链这类复杂机械设备,滤波频带的选择对包络分析结果影响很大,甚至可能得出错误的诊断结果。针对这一问题,提出一种称为“移动滤波包络谱图”的改进方法,利用中心频率移动的窄带滤波器对振动信号进行分频段窄带滤波处理,求出每段滤波信号的幅值包络谱,构成一组随窄带滤波中心频率变化的“频-频”域的包络谱阵,定义为“移动滤波包络谱图”。实例分析表明,移动滤波包络谱图可以清楚地区分正常状态和故障状态,直截了当地反映故障引起的振动信号各个频段的变化。 (3)齿轮、轴承等旋转机械零部件故障造成振动信号中随机成分的变化,对信号的循环平稳特性产生影响,因此可以用循环平稳特性分析中随机成分的变化反映故障。但是振动信号中随机成分的能量一般较低、分布频率范围较宽,为此提出利用谱相关密度三维对数等高图显示故障引起的信号随机成分的变化进行故障特征提取。通过试验轴承和直驱式风电机组轴承的实测振动信号对比分析,表明轴承故障产生振动信号的变化可以用对数谱相关函数图清楚地反映出来。以谱相关函数共振区切片低频部分(解调成分)的平均值作为特征值,由于消除了非循环平稳成分的影响,对于故障状态引起的变化更加敏感,有助于提高故障轴承诊断的准确性。
[Abstract]:Vibration monitoring is one of the main technology of condition monitoring and fault diagnosis of machinery, is widely used in many industries. Early wind farm construction in China for cost reduction, is generally not equipped with wind turbine vibration monitoring system. With the equipment running time the continuation of the wind turbine drive train failure rate and maintenance the problem of high cost gradually prominent, wind power companies began to pay attention to the wind turbine operation condition monitoring and fault diagnosis, especially vibration monitoring guidelines > recommended national energy industry standard NB/T31004-2011< wind power generation unit of National Energy Bureau issued in 2011, for the development of wind turbine vibration monitoring and fault diagnosis technology to a strong role in promoting.
The wind turbine transmission chain is mainly composed of a drive shaft and bearing, gear box and other rotating parts is connected with a wind wheel and a generator, a key component of energy conversion and transfer process. Compared with the transmission chain of general mechanical equipment, the load state of wind turbine transmission chain operating conditions, environmental conditions the structure and layout are complicated, resulting in gear transmission chain, bearings and other major components of the failure mechanism and failure mode of development has the particularity, the failure rate is higher than that of other similar equipment industry, the actual service life is far lower than the design life. Therefore, the common problems in the operation characteristic and fault diagnosis of wind turbine drivetrain. Research, new theory and technology solutions, to improve the accuracy of fault diagnosis and prediction, it is very important to ensure the healthy operation of the wind turbine equipment.
The wind turbine transmission chain is machine, electricity, liquid coupling of complex structure, possible fault location, fault excitation from source to vibration monitoring points through different transmission attenuation and mixing, vibration monitoring signal is often complex convolution mixing different excitation sources and transfer paths, with background noise large, non-stationary and nonlinear characteristics, in many practical situations, using vibration signal analysis method is difficult to give accurate information to reflect the classical fault features and influence the accurate analysis to the fault diagnosis and application limit automatic fault diagnosis and prediction technology. Therefore, according to the particularity of the gear box and the supporting bearings of wind power the unit in the transmission chain, the improved method of basic research in the classical analysis method on vibration signal, improve fault feature analysis and extraction of function and effect, is to achieve accurate fault diagnosis technique In this context, this paper discusses the improvement of several classical vibration signals analysis, and validates the measured signals of doubly fed and direct drive wind turbines respectively. The main contents and conclusions are as follows:
(1) the important properties of the cepstrum has a "homomorphic filtering", can realize the separation of convolutive mixed signals. Homomorphic filtering properties using cepstrum, presents a method of fault feature extraction of gearbox, the vibration signal detected is the result of convolution vibration excitation source and the transfer process, there different characteristics of the two signal components in cepstrum domain. Accordingly, the gearbox vibration signal processing in bandpass filtering in cepstrum domain, the excitation source is an inherent component and structure vibration components were separated, and then the spectrum reconstruction of structure components with low frequency resonance properties, changes reflect the structural characteristics caused by fault with the reconstruction of the spectrum, so as to reveal the trend of fault occurrence and development. According to the analysis of examples of wind turbine gearbox and bearing vibration, the extracted value is by operation The effect is small, and the fault state can be clearly characterized under various operating conditions, and it can also reflect the development trend of the fault.
(2) narrow-band envelope Hilbert transform analysis with vibration signal reflects the structure resonance frequency band is narrow band filter based on gain amplitude envelope. The main fault information reflecting the problem of the method is not to determine the frequency of narrowband filter, especially for the wind turbine drivetrain this kind of complex mechanical equipment, the filtering band selection on the envelope the analysis results are very different, and may get the wrong diagnosis results. To solve this problem, we propose an improved method called "mobile filtering envelope spectrum", the vibration signal frequency narrowband filtering processing using narrowband filter with the center frequency of the mobile, calculate the amplitude envelope of each section of the filtered signal spectrum, form an envelope a group with narrowband filter center frequency changes in the frequency - domain frequency spectrum array, defined as "mobile filtering envelope spectrum. Example analysis shows that the movement of the filter envelope spectrum profile In order to clearly distinguish the normal state and the failure state of the region, the change of each frequency band of the vibration signal caused by the fault is directly reflected.
(3) gears, bearings and other parts of rotating machinery fault caused by random changes in vibration signals, the influence of the cyclostationarity of signals, it can reflect the fault with changes of circulating random component stationarity analysis. But the random components in the vibration signal energy is generally low, the distribution of a wide range of frequencies, is proposed using spectral correlation density 3D contour map shows logarithmic change signal caused by the fault of the random components for fault feature extraction. Through the test of bearing and direct drive wind turbine bearing vibration signal analysis, indicating that the change of bearing fault vibration signals can be clearly reflected by logarithmic graph spectral correlation function. Based on the spectrum the correlation function (low frequency resonance demodulation section components) the average value as the characteristic value, due to the elimination of the influence of non cyclostationary components, for fault state The resulting changes are more sensitive and help to improve the accuracy of fault bearing diagnosis.

【学位授予单位】：华北电力大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TM315;TH165.3

【参考文献】