风力发电机传动系统故障诊断的时频综合分析方法研究

发布时间：2018-05-08 01:12

本文选题：风力发电机 + 故障诊断　；参考：《电子科技大学》2014年硕士论文

【摘要】：风能是储存量丰富的可再生洁净能源,是人类最早利用的能源之一。在利用它转换为电能的过程中,由于不产生有害气体和废料,也就是说它不污染环境,因此受到世界各国政府的广泛重视。随着风能技术的快速发展和日趋完善,风力发电机的可靠性越来越高,但是在风力发电系统的高速发展的同时,风力发电机故障问题也收到了极大的关注,如风机自身的轴承磨损、齿轮断齿、轴偏心等常见故障,都可能造成风机毁坏,从而降低经济效益,鉴于这些,风机故障诊断已经逐渐成为风力发电发展中的重要研究内容。风力发电机由于长时间地工作,一些部件之间的摩擦就会使得仪器的部件老化或者磨损,这无疑降低了发电机的寿命,降低了发电机的工作效率,而最常见的故障就是齿轮断齿和轴承的磨损,而对这些故障的故障诊断方法有直接观察法、振动和噪声检测法、无损检测法、磨损残余物检测法、机器性能参数检测法等。为了对这些故障进行高效地诊断,传统的直接观察法判断已经不足以满足风力发电机的发展和工作要求,所以一般是利用振动检测法,振动信号是机械设备状态信息的载体,包含了丰富的故障特征信息,故障诊断就是通过各种信号处理方法,把隐藏在振动信号中的有意义的特征信息提取出来,实现对设备的诊断。因此采用风力发电机振动信号判断其故障是一种可靠的方式,而对振动信号最合适的处理方式就是信号的时频分析方法,传统的时频分析方法如短时傅立叶变换、Wigner-Ville分布等,它们分别存在着窗效应和交叉项的问题且都不是自适应的。本文研究了一种改进的Hilbert-Huang变换(HHT),并将这改进的时频分析方法运用到风力发电机故障信号处理中。Hilbert_Huang变换通过经验模态分解(EMD)的方式将信号分解成为有限个固有模态(IMF),每一个固有模态都是一个稳态的信号,因此可以对每一个固有模态进行Hilbert变换,得到了信号的Hilbert谱,从而得到了信号的特征,结合模糊神经网络系统,对故障的类型进行更加准确和方便地识别。但是传统的Hilbert_Huang变换存在着端点飞翼和终止条件的问题,本文针对这两个缺陷进行了一系列的研究,最终形成了了一个种改进的HHT分析方法,利用这种改进的HHT分析方法获得了发电机的振动信号的时频特征,但是HHT方法只是得到信号的时频特征,但是并没得到最后的故障诊断结果,在目前的研究中,模糊神经网络已经得到了较快发展,并且显示出了强大的优势,它是一种新的诊断和识别技术,它将模糊逻辑推理的强大结构性知识表达能力和神经网络的强大自学习能力结合为了一体,本文将HHT分析得到的时频特征结合了模糊神经网络进行故障类型的诊断和识别,获得了很好的效果。
[Abstract]:Wind energy is a renewable clean energy with abundant storage, and it is one of the earliest energy sources used by human beings. In the process of using it to convert electric energy, because it does not produce harmful gas and waste, that is to say, it does not pollute the environment, so the governments all over the world pay more attention to it. With the rapid development and improvement of wind energy technology, the reliability of wind turbine is becoming more and more high. However, with the rapid development of wind power system, the problem of wind turbine fault has also received great attention. Such as bearing wear, gear tooth breaking, shaft eccentricity and other common faults may cause fan damage, thus reducing economic benefits. In view of these, fan fault diagnosis has gradually become an important research content in the development of wind power generation. Because wind turbines work for a long time, friction between some parts will cause the parts of the instrument to age or wear, which undoubtedly reduces the life of the generator and reduces the efficiency of the generator. The most common fault is the wear of gear broken teeth and bearing. The fault diagnosis methods of these faults include direct observation method, vibration and noise detection method, nondestructive testing method, wear residue detection method, machine performance parameter detection method and so on. In order to diagnose these faults efficiently, the traditional direct observation method is not enough to meet the development and working requirements of wind turbine, so the vibration signal is the carrier of mechanical equipment status information. Fault diagnosis is to extract the meaningful feature information hidden in the vibration signal through various signal processing methods to realize the diagnosis of the equipment. Therefore, it is a reliable way to use vibration signal of wind turbine to judge its fault, and the most suitable way to deal with vibration signal is time-frequency analysis method, traditional time-frequency analysis method such as short time Fourier transform Wigner-Ville distribution, etc. They have the problem of window effect and crossover, respectively, and they are not adaptive. In this paper, an improved Hilbert-Huang transform is studied, and the improved time-frequency analysis method is applied to wind turbine fault signal processing. Hilbert Huang transform decomposes the signal into finite inherent modes by empirical mode decomposition (EMD). IMF, each inherent mode is a steady-state signal, Therefore, the Hilbert transform can be carried out for each inherent mode, and the Hilbert spectrum of the signal can be obtained, thus the characteristics of the signal can be obtained, and the fault type can be identified more accurately and conveniently with the combination of the fuzzy neural network system. However, the traditional Hilbert_Huang transform has the problems of terminal wing and termination condition. In this paper, a series of research on these two defects has been carried out, and an improved HHT analysis method has been developed. The improved HHT analysis method is used to obtain the time-frequency characteristics of the generator vibration signals, but the HHT method only obtains the time-frequency characteristics of the signals, but the final fault diagnosis results are not obtained. Fuzzy neural network (FNN) has been developed rapidly and has shown strong advantages. It is a new diagnosis and recognition technology. It combines the strong structural knowledge expression ability of fuzzy logic reasoning with the powerful self-learning ability of neural network. In this paper, the time-frequency features obtained by HHT analysis are combined with fuzzy neural network to diagnose and identify fault types. Good results have been achieved.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM315

【参考文献】