科研之友
微信
新浪微博
Facebook
分享链接
摘要
Vibration signals have been widely used to detect gear faults in machines in various industrial applications. However, stator current signals from the drive motor have recently become an alternative and promising tool for detecting and diagnosing the existence and occurrence of gear faults, due to their higher reliability, lower cost and easier distant monitoring ability than traditional vibration-based fault detection methods. Therefore, this paper combines empirical mode decomposition (EMD), fast independent component analysis (FastICA) and a sample entropy measure to propose a hybrid feature extraction methodology for gear fault detection, which is used to analyse the stator current signals of the drive motor. First, the stator current signals obtained from an induction motor are decomposed by EMD into several intrinsic mode functions (IMFs). The signal-to-noise ratio of the stator current signals can be enhanced by removing the IMFs in high-frequency bands. Second, in order to eliminate information redundancy among the IMFs and improve the accuracy of fault detection, the FastICA approach is applied to the selected IMFs to extract independent components. Finally, the sample entropy of the independent components is calculated to quantitatively characterise the differences between healthy and faulty gears and then identify a gear pitting fault. The proposed method is verified by experiments on a real gearbox under different motor rotating speeds. The results demonstrate that the proposed hybrid feature extraction method can provide a more effective and efficient approach to gear pitting fault detection.
- 出版日期2014-6
- 单位燕山大学
