风力发电机齿轮箱故障特征提取及分类方法的研究

发布时间：2018-03-30 10:49

本文选题：齿轮箱　切入点：EEMD-ICA　出处：《燕山大学》2013年硕士论文

【摘要】：风力以其清洁可再生蕴藏量大分布广等优点引起了人们的重视。现在风力资源主要是是用于风力发电，风力发电机作为连接机械能与电能交换的纽带在风力发电中起着至关重要的作用，如果发电机出现问题，直接影响电力系统的正常运行。本文主要研究风力发电机齿轮箱的机械故障检测方法。首先，介绍了风力发电的工作原理以及组成部分。针对于风力发电机的齿轮箱的常见故障做了说明，并阐述了齿轮箱的振动机理。其次，对信号进行了一系列的处理，主要包括有信号的消噪部分，信号的分解部分，以及特征量的提取。在经验模态分解的基础之上提出了EEMD-ICA相结合的滤波方法，使其滤波效果更加良好；本文采用局部均值分解，求取分量的多重分形谱与近似熵相结合提取特征量可以更加全面的表示信号的特征。接着，将特征量作为模糊C聚类的输入量，进行模式识别，得到不同故障的分类，为机械故障诊断奠定基础。最后，，将该方法用于实验平台采集的数据，通过对数据的分析验证了以上所提方法的有效性。EEMD-ICA滤波可以使信号达到很好的滤波效果；多重分形与近似熵相结合可以对信号做出较全面的定量分析；模糊C聚类可以得到较好的聚类效果。
[Abstract]:Wind with its cleanliness? Renewable? Big reserves? Nowadays, wind power resources are mainly used for wind power generation. Wind turbines play an important role in wind power generation as a link between mechanical energy and electric energy exchange. If there is a problem with the generator, it will directly affect the normal operation of the power system. This paper mainly studies the mechanical fault detection method of the gearbox of the wind turbine. Firstly, the working principle and components of wind power generation are introduced, the common faults of the gearbox of wind turbine are explained, and the vibration mechanism of the gearbox is expounded. Secondly, a series of signal processing, including signal de-noising, signal decomposition and feature extraction, are presented. Based on empirical mode decomposition, a filtering method based on EEMD-ICA is proposed. In this paper, the multifractal spectrum of the component and the approximate entropy can be used to extract the features of the signal. Then, the feature quantity is used as the input of fuzzy C clustering, and the classification of different faults is obtained by pattern recognition, which lays a foundation for mechanical fault diagnosis. Finally, the method is applied to the data collected by the experimental platform. The validity of the proposed method. EEMD-ICA filter can make the signal achieve a good filtering effect through the analysis of the data. The combination of multifractal and approximate entropy can make a more comprehensive quantitative analysis of the signal, and fuzzy C clustering can get a better clustering effect.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TM315;TH165.3

【参考文献】