【作者】 马涛；

【导师】 陈长征；

【作者基本信息】 沈阳工业大学 ， 机械设计及理论， 2013， 硕士

【摘要】 风能作为一种可持续的清洁能源，许多国家已经把它作为可持续发展战略的重要能源。风力发电机是将风能转换成电能的主要设备，而齿轮箱作为风力发电机的关键机械系统，其健康状态对风力发电机的正常运行起着至关重要的作用。齿轮箱在运行过程中常见的故障有齿轮断齿、齿根裂纹、轴承的失效、轴磨损等，严重影响着风力发电机的正常运行。齿轮箱产生的故障信号通常是非平稳的振动信号，而且噪声信号常常把故障振动信号的特征淹没，尤其是同时出现多种故障，更不易实现故障状态识别。针对风力发电机齿轮箱中齿轮和轴承振动信号的振动机理及振动特点，采用BP神经网络作为故障分类的方法，通过对经验模式分解获得的本征模式函数进行希尔伯特变换，降低信号干扰，从而进行故障状态识别、为实现预测性维护齿轮箱奠定了基础。本文主要从以下几方面进行研究：（1）介绍了风力发电机故障诊断技术，研究齿轮箱振动故障的特点及故障的类型，分析几种故障振动信号的特征，进行信号处理。（2）采用一种基于希尔伯特-黄变换边际谱分析的齿轮箱故障诊断方法实现对齿轮箱单一故障与复合故障的诊断。该方法通过对经验模式分解获得的本征模式函数进行希尔伯特变换，得到边际谱，采用谱分析的方式实现齿轮箱故障的判定。（3）采集某风场数据，应用Matlab进行信号分析，提取故障特征频率。对齿轮箱健康状态、主轴磨损状态、轴承外圈点蚀状态和高速轴齿根裂纹状态进行分类识别。（4）提出了BP神经网络对齿轮箱状态进行预测。根据神经网络算法建立模型实现齿轮箱故障的类型识别，齿轮箱振动信号在Matlab平台上处理结果表明神经网络系统可以实现齿轮箱状态趋势预测。更多还原

【Abstract】 Wind energy, as a sustainable clean energy, has been used as an important energysource for the sustainable development strategy in many countries. Wind turbine is theequipment that converts wind energy into electricity. Gearbox, as the critical mechanicalsystem of wind turbine, determines the stability of wind turbine. There are some commonfailure of the gear box during operation, such as gear broken，surface pitting，bearingfailure，shaft bending and so on，which seriously affect the function of wind turbine.The fault signal generated by the gear box is usually non-stationary vibration signaland also usually disturbed by other noise generated by wind turbine. It will more difficultto achieve the recognition of fault state when several failures occur in same time.The thesis based on the vibration mechanism and vibration characteristics of the gearand bearing of wind turbine, with the BP neural network, the intrinsic mode functionsthrough empirical mode decomposition is obtained using the method of Hilbert transform,and the interference is reduced. At last, the foundation of the fault state recognition andpredictive maintenance of gearbox were made. The research as follows:(1)The thesis introduces the wind generator fault diagnosis technology, the gearvibration fault characteristics, fault types and the analysis of the characteristics of severalkinds of fault vibration signal, signal processing.(2)Using a kind of gear fault diagnosis method based on Hilbert-Huang transformingmarginal spectrum analysis and achieving gear box single fault and composite faultdiagnosis. The intrinsic mode functions through empirical mode decomposition is obtainedusing the method of Hilbert transforming is got marginal spectrum, which analysis methodto realize gearbox fault judgments.(3)Fault characteristic frequency is extracted by collecting a wind field data and thenanalyzing the signal data by Matlab. Gearbox health states, main wear state, bearing outer punctuate corrosion condition and high speed shaft root crack state are classified andidentified.(4)The state of gear box is forecast by putting forward the BP neural network.According to the neural network algorithm model, the gear box fault type recognition isachieved. The results of the gear vibration signal in the Matlab platform processing showsthat the neural network system can realize gear box state trend prediction.更多还原