国外文献 > 1.轴承 > 7. 智能方法与计算 > 神经网络
关键词搜索：
| 标 题： |
A Fault Diagnosis Method for Rolling Bearing Combining Signal Difference and Coarse Graining
结合信号差分与粗粒化技术的滚动轴承故障诊断方法 |
| 作 者： |
Yu Mingyue
Li Yongpeng
Ge Xiangdong
Li Zhaohua |
| 关键词： |
Fault diagnosis [故障诊断]
Differential signal [ 差分信号]
Support vector machine [ 支持向量机]
Rolling bearing [ 滚动轴承] |
| 出版年： |
2024 |
| 期刊名： |
Journal of Failure Analysis and Prevention
失效分析与预防杂志 |
| 年卷期： |
2024, vol.24, no.4 |
| 摘 要： |
To precisely determine the type of bearing fault, the paper has proposed the solution by combining first-order difference of signals and coarse-graining. When a bearing fault occurs, vibration signals are often accompanied by impact features and meanwhile, greater difference among adjacent points of signals corresponds to more evident impact features. Compared with original vibration signals, the first-order difference of signals works better in measuring this kind of impact feature. Therefore, the solution proposed in the paper takes first-order difference in place of original signals and combines with coarse-graining algorithm for subsequent treatment. Compared with average value of signals, the variance of signal can represent the discreteness of signal which describes the impact feature of bearing fault from the other side. Therefore, the solution has coarse-graining operation for difference signals in improved coarse-graining method with variance as criterion to further highlight impact features of rolling bearing. Secondly, a feature set has been established to comprehensively represent fault information of bearing with the average of coarse-graining sequence as feature parameter. Finally, the type of bearing fault is identified according to the fault feature set and support vector machine (SVM). To verify the effectiveness and universality of solution proposed by the paper, a comparison is given to proposed solution and classical method according to dataset of faults of rotor and rolling bearing, and public dataset. The result indicates that with different rotate speeds and types of faults, the proposed solution is more precise to identify the type of bearing faults; in two different datasets, the average identification rate of the proposed solution is 96.43% (aeroengine–rotor–rolling bearing fault dataset) and 100% (public dataset) for unknown samples.
为准确判断轴承故障类型，本文提出了一种结合信号一阶差分与粗粒化处理的方法。当发生轴承故障时，振动信号通常伴随冲击特征，同时信号相邻点之间的差异越大，对应的冲击特征越明显。与原始振动信号相比，信号的一阶差分在检测此类冲击特征方面表现更优。因此，本文提出的方案以信号的一阶差分替代原始信号，并结合粗粒化算法进行后续处理。相较于信号的均值，信号的方差能够反映信号的离散程度，从另一角度描述轴承故障的冲击特征。因此，该方案采用改进型粗粒化方法，以方差作为判据对差分信号进行粗粒化操作，进一步突出滚动轴承的冲击特征。其次，建立了一组特征集，以粗粒化序列的平均值作为特征参数，全面表征轴承故障信息。最后，根据故障特征集和支持向量机（SVM）识别轴承故障类型。为验证本文所提方法的有效性和普适性，对所提方法与经典方法在转子及滚动轴承故障数据集和公开数据集上的表现进行了比较。结果表明，在不同转速和故障类型下，所提方法在识别轴承故障类型方面具有更高的精确度。在两个不同的数据集中，所提出方法的平均识别率分别为96.43%（航空发动机-转子-滚动轴承故障数据集）和100%（公共数据集），对于未知样本。 |
| 中文内容为网络AI翻译，专业术语未必准确，仅供参考。 |
京公网安备 11010202009788