基于改进EEMD的风电机组行星齿轮箱故障诊断研究

发布时间：2018-11-03 11:23

【摘要】：我国作为一个能源需求大国,对能源的需求与日俱增。但我国的能源结构处于欠合理状态,主要表现为对于化石能源的依赖严重,清洁能源占比不高等问题。随着化石能源的枯竭、环境恶化等问题的出现,都要求我国逐步发展清洁能源来改变传统的能源结构。在众多清洁能源中,风能作为其中最具代表性的一种,凭借着分布广泛、商业化程度高、技术成熟等优势正在越来越多的发挥着重要作用。但风电机组工作环境恶劣,经常面临风速不稳定、内外环境温差大、载荷多变等问题。不同类型的风电机组也将面临不同的恶劣环境,如海上风电机组所处环境空气湿度大、盐分高,机组中零部件易受到腐蚀;陆上风电机组面临的最大环境问题是空气中沙尘大,当机组密封条件不佳时,沙尘进入机组极易造成齿轮损坏等问题,众多因素导致了风电场运维成本持续居高不下。据估计,在风电机组的运行寿命周期内,运维费用是发电总成本的重要组成部分,约占总成本的25%~30%。同时,对海上风电机组的运行统计中发现,50%的停运时间是由齿轮箱故障造成的。根据以上数据可以看出,对风电机组行星齿轮箱运行状态做出及时的识别与诊断,具有极大实际应用意义。本文主要研究了风电机组行星齿轮箱故障的主要原因及其故障检测的有效方法:1)分析了风电机组行星齿轮箱中不同类型故障出现的主要原因,总结了不同类型的故障特征,并针对不同故障特征提出了对应的运行维护方法,提高了机组运行的可靠性。2)提出一种基于改进EEMD的自适应信号分解方法。可以针对不同信号自适应给出不同的分解参数,在实际应用中一定程度的解决了传统EEMD分解过程中的模态混叠问题、提高了计算速度、改善了分解效果。达到了信号自适应分解的目的。3)利用提出的改进EEMD方法实际信号进行分解,然后使用单重分形维数提取经改进EEMD分解后各个分量中的分形特征,通过对比信号特征实现了对行星齿轮箱故障的实时有效的诊断。4)利用多重分形维数谱与支持向量机的结合方法,实现了对行星齿轮箱在不同转速情况下的故障的诊断,进一步证实了分形维数对于信号特征具有良好的提取能力。同时证明了支持向量机对于信号的分类效果良好。5)最后通过使用单重分形维数提取经过改进EEMD分解后的各个分量的分形特性,然后将提取的得到的信号特征作为支持向量机的输入向量进行信号分类,实现了对实际故障信号的有效诊断。
[Abstract]:As a large country of energy demand, China has a growing demand for energy. However, the energy structure of our country is in an unreasonable state, which is mainly due to the heavy dependence on fossil energy and the low proportion of clean energy. With the depletion of fossil energy and the appearance of environmental deterioration, China needs to develop clean energy gradually to change the traditional energy structure. Among the many clean energy sources, wind energy, as one of the most representative, is playing an important role with the advantages of wide distribution, high degree of commercialization, mature technology and so on. However, wind turbine often faces some problems such as unstable wind speed, large temperature difference between inside and outside environment, variable load and so on. Different types of wind turbine units will also be faced with different adverse environment, such as offshore wind turbine units in high air humidity, high salinity, unit components are vulnerable to corrosion; The biggest environmental problem faced by onshore wind turbines is the large dust in the air. When the sealing conditions of the units are not good, the sand dust entering the units is easy to cause gear damage and so on. Many factors cause the operation and maintenance costs of the wind farms to remain high. It is estimated that the cost of operation and maintenance is an important part of the total cost of power generation in the operational life cycle of wind turbine, accounting for about 2530% of the total cost. At the same time, it is found that 50% of the outage time is caused by gearbox failure. It can be seen from the above data that it is of great practical significance to identify and diagnose the running state of planetary gearbox of wind turbine unit in time. In this paper, the main causes of planetary gearbox faults of wind turbine and the effective methods of fault detection are studied. 1) the main causes of different types of faults in planetary gearboxes of wind turbines are analyzed, and the characteristics of different types of faults are summarized. According to different fault characteristics, the corresponding operation and maintenance methods are proposed to improve the reliability of unit operation. 2) an adaptive signal decomposition method based on improved EEMD is proposed. Different decomposition parameters can be given according to different signal adaptations. In practical application, modal aliasing in the traditional EEMD decomposition process is solved to a certain extent, and the calculation speed is improved and the decomposition effect is improved. The purpose of adaptive signal decomposition is achieved. 3) the improved EEMD method is used to decompose the actual signal, and then the fractal features of each component after the improved EEMD decomposition are extracted by using the single multifractal dimension. The real-time and effective fault diagnosis of planetary gearbox is realized by comparing the signal features. 4) the fault diagnosis of planetary gearbox under different rotational speeds is realized by the combination of multifractal dimension spectrum and support vector machine. It is further proved that fractal dimension has a good ability to extract signal features. At the same time, it is proved that SVM has a good effect on signal classification. 5) finally, the fractal characteristics of each component after improved EEMD decomposition are extracted by using single multifractal dimension. Then, the extracted signal features are classified as input vectors of support vector machine (SVM), and the effective diagnosis of actual fault signals is realized.
【学位授予单位】：上海电力学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】