声发射技术在超低速轴承故障诊断中的应用研究

发布时间：2018-12-12 09:55

【摘要】：滚动轴承作为旋转机械设备中最为常用的关键零部件之一,其运转状况往往直接关系到整台设备的安全稳定运行,一旦产生故障,将会极大的影响机械设备的生产安全和效率。因此,对滚动轴承的损伤状态进行监测诊断就显得尤为重要。低速重载轴承由于其运转和本身结构的复杂特殊,对于这类轴承的损伤状态进行监测异常困难。随着机械制造行业的快速发展,低速重载轴承的实际应用范畴也越来越广泛,并且这类轴承一般都安装在大中型的旋转机械设备中,一旦产生损坏造成停机,其维修更换需要大量的人力物力财力,因此提前监测这类轴承的损伤状态能够避免停机事故,获得较大的经济效益。声发射技术(AE)是一种新型的动态实时监测技术,与传统的检测技术相比,声发射信号对动态缺陷敏感、频带较宽,检测效率高,可以及时的发现低速重载滚动轴承的早期损伤,对于旋转机械设备中轴承的保养和维修具有重要的工程应用价值。本文以声发射检测技术为手段,通过搭建实验台来模拟低速重载轴承的运行状态,对轴承预制不同位置和大小的人工缺陷,采集不同损伤状态轴承的声发射信号,对其在低速轴承故障诊断和损伤状态监测的可行性进行了理论和实验研究。完成的主要工作和成果有:借助试验台采集不同损伤状态的轴承声发射信号,分别采用小波分析和小波包分析对信号进行分析处理,通过提取各频带所占能量百分比,得出相比小波分析,小波包分析能够提取到轴承故障声发射信号产生的主要频带,提取的能量较高的频带与频谱图高幅值频带相一致。并比较了小波尺度谱和STFT谱对低速轴承AE信号中的特征提取性能,结果表明小波尺度谱对于非平稳声发射信号的时间分辨率较高,而STFT谱相比小波尺度谱对于非平稳信号的频率分辨率较高,因而可以将小波尺度谱和STFT谱相结合用于低速轴承故障特征提取。针对低速轴承故障特征微弱,易被噪声淹没,提出了结合能量熵和集合经验模态分解(EEMD)进行低速轴承故障诊断,并提出基于相关系数法和方差贡献率法筛选有效本征模态分量。通过实验结果表明,采用互相关系数和方差贡献率能够筛选有效的IMF分量,提取的有效IMF能量熵能够很好的表征低速轴承的损伤缺陷变化。并对比了支持向量机和BP神经网络对低速轴承的故障类型的分类识别效果,得出针对低速轴承小样本数据支持向量机的识别准确率要高于BP神经网络。
[Abstract]:Rolling bearing is one of the most commonly used key parts in rotating machinery. Its running condition is often directly related to the safe and stable operation of the whole equipment. Once failure occurs, it will greatly affect the production safety and efficiency of mechanical equipment. Therefore, it is very important to monitor and diagnose the damage state of rolling bearing. It is very difficult to monitor the damage state of low speed heavy load bearing because of its complex structure and operation. With the rapid development of machinery manufacturing industry, the practical application of low-speed and heavy-duty bearings is becoming more and more extensive, and this kind of bearings are generally installed in large and medium-sized rotating machinery equipment. The maintenance and replacement need a lot of manpower and financial resources, so monitoring the damage state of this kind of bearing in advance can avoid the downtime accident and obtain greater economic benefit. Acoustic emission technology (AE) is a new dynamic real-time monitoring technology. Compared with traditional detection technology, acoustic emission signal is sensitive to dynamic defects, wide frequency band and high detection efficiency. The early damage of low speed heavy load rolling bearing can be found in time. It has important engineering application value for the maintenance and repair of bearing in rotating machinery and equipment. In this paper, the acoustic emission testing technology is used to simulate the running state of low-speed and heavy-load bearing by building an experimental bench. The acoustic emission signals of bearings with different damage states are collected for the artificial defects in different positions and sizes of prefabricated bearings. The feasibility of fault diagnosis and damage monitoring of low speed bearing is studied theoretically and experimentally. The main work and achievements are as follows: the acoustic emission signals of bearings with different damage states are collected by means of the test bed, the signals are analyzed and processed by wavelet analysis and wavelet packet analysis, and the percentage of energy occupied by each frequency band is extracted. Compared with wavelet analysis, wavelet packet analysis can extract the main frequency band of acoustic emission signal of bearing fault, and the frequency band with higher energy is consistent with the high amplitude frequency band of spectrum diagram. The feature extraction performance of wavelet scale spectrum and STFT spectrum for low speed bearing AE signal is compared. The results show that wavelet scale spectrum has higher time resolution for non-stationary acoustic emission signal. Compared with wavelet scale spectrum, STFT spectrum has higher frequency resolution for non-stationary signal, so wavelet scale spectrum and STFT spectrum can be combined to extract fault feature of low-speed bearing. Since the fault characteristics of low speed bearing are weak and easily submerged by noise, a low speed bearing fault diagnosis based on energy entropy and set empirical mode decomposition (EEMD) is proposed. Based on the correlation coefficient method and the variance contribution rate method, the effective intrinsic modal components are selected. The experimental results show that the effective IMF component can be selected by using the cross-correlation number and variance contribution rate, and the extracted effective IMF energy entropy can well characterize the damage and defect change of low-speed bearing. The classification and recognition effects of support vector machine and BP neural network on low speed bearing fault types are compared. It is concluded that the recognition accuracy of support vector machine for small sample data of low speed bearing is higher than that of BP neural network.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.3

【参考文献】