基于改进集成学习算法的风机齿轮箱故障诊断与状态监测研究

发布时间：2018-10-05 13:29

【摘要】：随着全球新能源技术的迅猛发展以及风电比重的不断提高,风机故障诊断与状态监测方法研究的重要性及紧迫性也越来越显著。本文以风机齿轮箱故障诊断与状态监测方法为研究重点,以提高风机齿轮箱的可靠性为目的,分别对风机齿轮箱齿轮、轴承以及油温进行研究。全文围绕集成学习与人工蜂群算法以及改进的人工蜂群算法的结合,对风机齿轮箱故障诊断和状态监测进行研究,并结合研究方法对风机故障诊断与状态监测系统进行软硬件设计,主要做了以下四个方面的工作。(1)研究了风机齿轮箱故障形成机理,利用在实验平台上采集的齿轮、轴承振动信号,首先通过小波包变换方法对振动信号降噪,对降噪后信号提取时域特征值,然后对时域信号进行快速傅立叶变换得频域信号,再提取频域特征值,最后将时域特征值和频域特征值归一化,为下面章节进行故障诊断打好基础。(2)对风机齿轮箱中的齿轮点蚀、齿轮断齿、轴承内圈损坏以及轴承外圈损坏进行诊断,提出基于人工蜂群算法的选择性神经网络集成算法(ABCSEN)的故障诊断方法,首先利用UCI数据集验证了本文提出的ABCSEN在诊断精度和效率上都优于GASEN和Bagging,然后利用齿轮箱历史故障数据训练ABCSEN得到故障诊断模型,最后利用新的故障数据测试故障诊断模型,结果显示模型诊断效果较好。(3)以风机齿轮箱油温数据为基础,研究齿轮箱状态监测的方法,首先提出改进的ABCSEN,即基于动态柯西蜂群算法的选择性神经网络集成算法(MCABCSEN),然后利用测试函数验证了改进算法性能的优越性,最后利用南方某风场7#风机齿轮箱油温数据以及人为拟合的齿轮箱故障油温数据对新算法进行训练和测试,结果显示新算法状态监测灵敏,能提前预警故障并及时提醒工作人员,防止损失进一步加大。(4)基于研究的故障诊断和状态监测的方法,搭建了风机齿轮箱故障诊断与状态监测系统,较为详细的讲解了系统硬件和软件的设计,最后对设计方案进行了简单的实验验证。
[Abstract]:With the rapid development of global new energy technology and the increasing proportion of wind power, the importance and urgency of fan fault diagnosis and condition monitoring are becoming more and more obvious. In order to improve the reliability of fan gearbox, this paper focuses on the fault diagnosis and condition monitoring method of fan gearbox, and respectively studies the fan gearbox gear, bearing and oil temperature. Based on the combination of integrated learning, artificial bee colony algorithm and improved artificial bee colony algorithm, the fault diagnosis and condition monitoring of fan gearbox are studied in this paper. Combined with the research method, the software and hardware design of fan fault diagnosis and condition monitoring system is carried out, mainly in the following four aspects. (1) the mechanism of fan gearbox fault formation is studied, and the gears collected on the experimental platform are used. The vibration signal of bearing is firstly de-noised by wavelet packet transform, then the time domain eigenvalue is extracted from the signal after denoising, then the frequency domain signal is obtained by fast Fourier transform of the time domain signal, and then the frequency domain characteristic value is extracted. Finally, the time-domain and frequency-domain eigenvalues are normalized to provide a good basis for fault diagnosis in the following chapters. (2) the pitting corrosion of the gears in the fan gearbox, gear tooth breaking, bearing inner ring damage and bearing outer ring damage are diagnosed. A fault diagnosis method of selective neural network ensemble algorithm (ABCSEN) based on artificial bee colony algorithm is proposed. Firstly, the UCI data set is used to verify that the ABCSEN presented in this paper is superior to GASEN and Bagging, in accuracy and efficiency. Then the ABCSEN is trained with the historical fault data of the gearbox to obtain the fault diagnosis model. Finally, the new fault diagnosis model is tested with the new fault data. The results show that the model has good diagnostic effect. (3) based on the oil temperature data of fan gear box, the method of monitoring the condition of gear box is studied. The improved ABCSEN, (selective neural network ensemble algorithm (MCABCSEN), based on dynamic Cauchy swarm algorithm) is proposed firstly. Then the superiority of the improved algorithm is verified by using the test function. Finally, the new algorithm is trained and tested by using the oil temperature data of the gear box of a certain wind field in the south and the fault oil temperature data of the gear box fitted artificially. The results show that the new algorithm is sensitive to state monitoring. It can warn the fault in advance and warn the staff in time to prevent the loss from increasing further. (4) based on the method of fault diagnosis and condition monitoring, a fault diagnosis and condition monitoring system for fan gearbox is built. The design of system hardware and software is explained in detail, and the design scheme is verified by experiments.
【学位授予单位】：上海电机学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】