【作者】 曲弋；

【导师】 陈长征；

【作者基本信息】 沈阳工业大学 ， 机械设计及理论， 2012， 博士

【摘要】 风能作为现今重要的新型清洁能源，在我国得到了大力开发，但是在大型风电设备的故障诊断和动态监测中逐渐暴露出来许多问题。对于风力发电机组的大型化，干扰强，特征信号微弱，湍流、流固耦合等作用的影响，使得传统的检测方法受到限制，不足以完成对风力发电机组的故障监测。作为风力发电机组中增速系统的齿轮箱是其重要部件之一。目前，齿轮箱的动力学分析通常集中于行星轮系上，行星轮系也是风力发电机组齿轮箱中的重要部件。本文建立了行星齿轮结合平行轴的斜齿圆柱齿轮和轴系组成的整个齿轮箱的动力学模型，分析风电齿轮箱的振动激励及振动特征，根据动力学模型建立了数学表达式。采用子空间迭代法求解了系统的模态，并用实验案例证明了算法的可行性。也证明了风力发电机组中的各主要部件在稳定运行时，振动特征有其典型特性和限制，当部件内部出现故障，其振动的振幅形式及频谱成分会发生某些变化，不同的缺陷和故障对应着不同的振动方式，振动信号能客观的反映风力发电系统齿轮箱的运行状态，因此利用振动信号对风力发电机组的齿轮箱进行监测和诊断，是发现故障的重要技术手段。传统的故障诊断方法缺乏严格的理论推理和数据证明。本文提出神经网络方法对风力发电机组齿轮箱故障诊断进行研究，使得风力发电机组齿轮箱输入轴轴承的故障诊断有较为明确的诊断结果。本文探索了声发射技术在风力发电机组部件故障诊断中的应用。在振动信号分析的基础上提出了声信号的采集和处理，由于声发射信号有别于振动信号，声发射检测方法的应用为故障诊断和特征识别提供了新的手段和途径。利用振声信号有针对性地提取故障的特征，发挥各自优势推进了风力发电机组的实时监测技术的发展。现阶段针对风力发电机组叶片的故障监测研究较为少见，本文以空气动力学理论为基础，研究风场的湍流模型并进行了流场仿真计算，得到叶片根部受力较大和叶片在绕Y轴的力矩较大，而研究Z轴方向的力和力矩没有实际意义的结论。建立了叶片与流场的流固耦合模型，叶片颤振物理模型。分析了叶片颤振稳定性的判据，得出判断叶片颤振的刚度阻尼和气动阻尼的计算公式。对真实叶片进行模态分析，证实颤振的存在及出现的频率范围，为风力发电机组叶片的故障诊断奠定力学基础。提出用声发射方法研究风力发电机组叶片的故障特征。由于声发射信号的复杂性和易于受到干扰性，借助小波尺度谱来提取声发射信号特征，提出了用小波尺度谱方法处理风力发电机组叶片的裂纹故障声发射信号。通过搭建声发射设备检测风力机叶片复合材料块的试验平台，采集了扩展裂纹与萌生裂纹的声发射信号，辨识了扩展裂纹和萌生裂纹并归纳出相关判据。更多还原

【Abstract】 Wind power is an important and clean energy, which is developed vigorously in china.But many problems of fault diagnosis and dynamic monitoring has been exposed. Becauseof the large-scale wind turbine, the interference, weak signal characteristics, the role ofturbulence, and fluid-structure interaction, traditional methods is restricted and insufficientto complete the monitoring of wind turbine failure.The dynamic analysis of the gearbox concentrated in the planetary gear train usually.The planetary gear train is also an important component of the wind turbine gearbox. Inthis paper, the dynamic model of the planetary gear combined with ordinary gear train andshafting is constructed. Vibration excitation and vibration characteristics of the windturbine gearbox are introduced, and mathematical expression based on the dynamic modelis built. The subspace iteration method is used to solve the system modal, then real caseprove the feasibility of algorithm. The vibration characteristics of major components in thestable operation have typical feature and limitations. When some parts have internal failure,the form of vibration signal amplitude and spectral components will change. Differentdefects and malfunctions correspond to different modes of vibration. Therefore, vibrationsignal can reflect the operating state of gearbox objectively. Traditional fault diagnosismethods lack rigorous theoretical reasoning and proving data. This paper presents thatneural network approach is used to study wind turbine gearbox fault diagnosis. It wasconducted that specific results of bearing fault diagnosis can be achieved.The application of acoustic emission(AE) in wind turbine component fault diagnosisis proposed. The AE signal is different from the vibration signal. Therefore, it enrich thefault diagnosis methods and characteristics identification methods. AE signal and vibrationsignal are applied in different components in pertinences respectively. Using theiradvantages can push forward the development of wind turbine monitoring and faultdiagnosis. Nowadays, fault monitoring for wind turbine blades is rare. This paper presents theflow field simulation based on aerodynamic theory to study the turbulence of the windfield model.It is arrived that the force at root and torque around Y-axis are large, and forceand moment of Z-axis is little. Fluid-solid coupling model of Blade and flow field andblade flutter model are established. Blade flutter stability criterion is studied, and judgmentformula of leaves flutter stiffness damping and aerodynamic damping are obtained.Through the calculating of data of real blades, the presence and the frequency range ofblade flutter are confirmed, which lay the foundation for fault diagnosis of the wind turbineblades.Extraction early crack characteristics with strong noise and identification differentkinds of cracks in wind turbine blades are important in wind turbine fault diagnosis.Experimental platform has been constructed to test a wind turbine blade that is made ofcomposite material, and to collect the AE signal of propagation crack and initiation crack.Wavelet scalogram was used to extract crack AE signal characteristics and identify thepropagation crack and initiation crack for its superior time-frequency analysis feature. Theresults reveal that the wavelet scalogram can extract nonlinear, non-stationary faultfeatures effectively, which is better than the wavelet analysis. It is also indicated thecriterion of propagation crack and initiation crack. Finally, it has been established the newcrack identification method of wind turbine blades based on AE and wavelet scalogram.更多还原