国外文献 > 1.轴承 > 4. 设计与分析 > 算法与优化
关键词搜索：
| 标 题： |
Application of Approximate Entropy in Fault Diagnosis and trend prediction of Rolling Element Bearing
近似熵在滚动轴承故障诊断与趋势预测中的应用 |
| 作 者： |
Hongfang Yuan
Ren Zhang
Peng Wang
Huaqing Wang
Jinji Gao |
| 关键词： |
Approximate entropy [近似熵]
Fault diagnosis [ 故障诊断]
Genetic algorithm [ 遗传算法]
Trend prediction [ 趋势预测] |
| 出版年： |
2013 |
| 会议名： |
2013 International Congress of Condition Monitoring and Diagnostic Engineering Management
2013年国际状态监测与诊断工程管理大会 |
| 届 次： |
|
| 摘 要： |
Feature extraction is the foundation of fault diagnosis and trend prediction of rolling bearing, and a good characteristic parameter could improve detection efficiency and forecast accuracy. Approximate Entropy (ApEn), as a measure reflecting signal regularity, has a positive effect on detecting fault and representing defect severity. But as traditional calculation method of ApEn needs reconstructing vectors and calculating distance between vectors repeatedly, which cause long computing time, so it is unsuited for on-line real-time equipment monitoring. This paper presents a fast approach to ApEn by constructing distance matrix of a time series, replacing the first two steps of traditional method. Then numerical simulation verifies the effectiveness of the improved method. The computing rate of modified method improves above six times, and is almost independent of parameters in calculation procedure that are crucial to traditional method. Meantime, the relationship between ApEn value and calculating parameters (e.g. computing dimensionality and tolerance value) is discussed, and their reasonable initial range is proposed. As defect of equipment develops, the regularity of fault vibration signal would decrease and its corresponding ApEn value would increase. But as the time of rapid deterioration approaches, ApEn value may decrease as the number of fault frequency components in analyzed signal increases. This characteristic of ApEn makes it suitable for forecasting fault developing trend. A load testing of rolling bearing with an artificial slot failure used for fault trend prediction based on genetic algorithm has approved the efficiency and high robustness of ApEn on identifying operation condition on rolling element bearing.
特征提取是滚动轴承故障诊断和趋势预测的基础，良好的特征参数能够提高检测效率和预测精度。近似熵（ApEn）作为一种反映信号规律性的指标，在故障检测和缺陷严重程度表征方面具有积极作用。然而，传统ApEn的计算方法需要反复重构向量并计算向量之间的距离，导致计算时间较长，因此不适用于在线实时设备监测。本文提出一种快速的ApEn计算方法，通过构建时间序列的距离矩阵，替代传统方法的前两个步骤。随后通过数值模拟验证了改进方法的有效性。改进后的计算速度提升了六倍以上，且几乎不受传统方法中关键计算参数的影响。同时，文中探讨了ApEn值与计算参数（如计算维数和容差值）之间的关系，并提出了合理的初始范围建议。随着设备缺陷的发展，故障振动信号的规律性会降低，相应的ApEn值则会增加。但当快速劣化时间临近时，随着分析信号中故障频率分量数量的增多，ApEn值可能下降。这种ApEn的特性使其适用于预测故障发展趋势。利用遗传算法对带人工槽缺陷的滚动轴承进行负载测试，并基于此预测故障趋势，验证了ApEn在识别滚动元件轴承运行状态方面的有效性与高鲁棒性。 |
| 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