Journal of Vibration Engineering & Technologies ( IF 2.7 ) Pub Date : 2019-08-15 , DOI: 10.1007/s42417-019-00167-4 Thiago H. G. Lobato , Roger R. da Silva , Ednelson S. da Costa , Alexandre L. A. Mesquita
Purpose
Since reliability and extended service life of rotating machinery are the industries´ major concerns, fault diagnosis systems are constantly being improved, especially by artificial intelligence methods. Current paper proposes a diagnostic method integrating stationary and non-stationary signal processing techniques, selection of multiple attributes, and classification by machine-learning algorithm. The technique was applied to a small number of measured signals.
Method
The integrated method uses the ensemble empirical mode decomposition (EEMD) (which handles nonlinear and non-stationary data) for signal processing, and the support vector machine (SVM) for the classification of the machinery condition with a small number of signals. Augmented data and feature selection with a genetic algorithm are used to improve the accuracy of the analysis.
Results and Conclusions
Evaluation was obtained by vibration signals from a rotor test rig with different types of faults. Experimental results showed that the proposed method successfully identifies the rotor´s faults with accuracy of 95.19%.
中文翻译:
使用EEMD,SVM和增强数据的旋转机械故障诊断的集成方法
目的
由于旋转机械的可靠性和使用寿命的延长是行业关注的重点,因此故障诊断系统正在不断改进,尤其是通过人工智能方法。目前的论文提出了一种诊断方法,该方法结合了平稳和非平稳信号处理技术,多种属性的选择以及通过机器学习算法进行分类的方法。该技术被应用于少量的测量信号。
方法
集成方法使用整体经验模式分解(EEMD)(处理非线性和非平稳数据)进行信号处理,并使用支持向量机(SVM)对少量信号进行机械状态分类。使用增强数据和遗传算法进行特征选择可提高分析的准确性。
结果与结论
通过来自具有不同类型故障的转子测试台的振动信号获得评估。实验结果表明,该方法成功地识别出转子故障,准确率达到95.19%。