Fault Diagnosis of Locomotive Wheel-bearing Based on Wavelet Packet and MCA

Abstract

Fault diagnosis of locomotive wheel-bearing is directly related to the locomotive performance and the safe operation of train. Owing to the fault signal of locomotive wheel-bearing being difficult to separate, the fault diagnosis method was proposed, which based on wavelet packets and morphological component analysis combined with the vibration signal characteristics of locomotive wheel-bearing. The simulation results show that the fault diagnosis of the locomotive wheel-bearing under low signal-to-noise ratio (SNR) case is achieved by wavelet packet and morphological component analysis. It provides a theoretical basis for the fault diagnosis and condition monitoring for the locomotive wheel-bearing. Introduction Wheel-bearing plays a vital role in running parts of locomotive. High speed and heavy load for the railway transportation industry have become an irresistible trend. As the key part of locomotive running gear, the condition of wheel-bearing is directly related to the performance of locomotive and the safety of train. Therefore, fault diagnosis on locomotive wheel-bearing is of great significance to prevent fault and ensure safe operation. Furthermore, it provides technical foundation for remote condition monitoring of locomotive running parts [1]. In the past decades, rolling bearing for fault diagnosis has gained more attention with the fast development of manufacturing industry. Nowadays, the blind source separation for fault diagnosis of rolling bearings has been applied to many fields [2-3]. The fault diagnosis for rolling bearing based independent component analysis has been achieved [4-5]. The characteristics of the ICA are that original signal has been decomposed into several independent components according to the principle of statistical independence. Each signal component is recovered only by the original signal to achieve the separation and the extraction of signal. However, the limitation of this method is that it has certain requirements for the original signal. The composite fault of rolling bearing was successfully distinguished by morphological component analysis. However, this method took more time [6]. The resonant component and impact component of rolling bearing was separated successfully with the morphological component analysis. But the analysis time was too long. The blind source signal was separated perfectly in combination with the wavelet packet and the variational Bayesian independent component analysis method. We have proposed a fault diagnosis method, which consists of wavelet packet transform and morphological component analysis, to separate blind source signal under low signal-to-noise ratio case in this paper. Firstly, the wavelet packet decomposition and reconstruction are used to reduce the noise of original signal. Then, the signal is separated by morphological component analysis. Finally, the pulse signal is demodulated with Hilbert transform and the fault position of locomotive wheel-bearing is identified according to the characteristic frequency. 2nd International Conference on Electrical and Electronic Engineering (EEE 2019) Copyright © 2019, the Authors. Published by Atlantis Press. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). Advances in Engineering Research, volume 185