Railway transport is one of the world’s oldest means of transportation over long distances, used for transposition of people and goods. Latest developments of other faster and reliable means of transports such as airplanes and roads have become a threat to the railway’s transportation. Thus, several improvements have been made such as the increase in train operation speeds to ensure they remain competitive in the transport sector. This has resulted to the trains’ wheelsets, which comprises the axle, axle bearing, and the wheels, to contact fatigue due to the different environmental and operating condition hence resulting to mechanical failure in most cases. For instance, axle bearing failure is the worst as it may also lead to damages of the other parts of train thereby endangering the railway transportation and safety.
Presently, the need for the axle bearing failure detection techniques has been considered to be the remedy in solving such problems. Although several methods of detection of the axle bearing failures such as acoustic analysis and temperature detection have been put to use, their work has been limited due to the several environmental factors and conditions. Therefore, the use of multi-scale morphological filters analysis of the bearing vibration signals is on high increase as it is reliable and less complicated.
A group of researchers at Southwest Jiaotong University in China: Associate Professor Yifan Li, Professor Jianhui Lin, and ProfessorDr. Jianxin Liu in collaboration with Professor Xihui Liang and Yuejian Chen, a PhD student at University of Alberta in Canada proposed a novel multi-scale morphological filter for real-time fault detection in the axle bearings. The signal processing scheme was feature selection based. Gray relational analysis was applied in the identification of the scale and diagnostic features for the multi-scale morphological filter (MMF). Their work is published in the research journal, Mechanical Systems, and Signal Processing.
The authors successfully detected all the three bearing faults simultaneously, that is pin roller fault, cage fault, and outer ring fault, using the feature selection MMF based feature selection MMF, unlike the MMF method based on the kurtosis which could only detect at most two bearing faults and MMF method based on the spectral kurtosis (SK). However, kurtosis based MMF and SK based MMF require less computation efficiency as compared to the feature selection based MMF. This was due to the fact that the procedure for selecting the features is time-consuming.
From the experimental results obtained, the proposed novel feature selection based multi-scale morphological filter qualifies for use in the railway transportation for detecting the axle bearing faults. Its gives better detection abilities than its counterparts, that is, SK kurtosis and Kurtosis SK based MMF. The filtering scale is obtained based on the feature indicators and the gray relational analysis. As a result, it is much simpler to get the train axle bearing vibration signal of the features caused by the bearing faults. Furthermore, the authors are optimistic that this study will advance the train transport industries in enhancing the operation safety and operational costs.
Li, Y., Liang, X., Lin, J., Chen, Y., & Liu, J. (2018). Train axle bearing fault detection using a feature selection scheme based multi-scale morphological filter. Mechanical Systems and Signal Processing, 101, 435-448.Go To Mechanical Systems and Signal Processing