Abstract
The behavioural diagnostics of bearings play an essential role in the management of several rotation machine systems. However, current diagnostic methods do not deliver satisfactory results with respect to failures in variable speed rotational phenomena. In this paper, we consider the Shannon entropy as an important fault signature pattern. To compute the entropy, we propose combining stationary wavelet transform and singular value decomposition. The resulting feature extraction method, that we call stationary wavelet singular entropy (SWSE), aims to improve the accuracy of the diagnostics of bearing failure by finding a small number of high-quality fault signature patterns. The features extracted by the SWSE are then passed on to a kernel extreme learning machine (KELM) classifier. The proposed SWSE-KELM algorithm is evaluated using two bearing vibration signal databases obtained from Case Western Reserve University. We compare our SWSE feature extraction method to other well-known methods in the literature such as stationary wavelet packet singular entropy (SWPSE) and decimated wavelet packet singular entropy (DWPSE). The experimental results show that the SWSE-KELM consistently outperforms both the SWPSE-KELM and DWPSE-KELM methods. Further, our SWSE method requires fewer features than the other two evaluated methods, which makes our SWSE-KELM algorithm simpler and faster.
Highlights
Diagnosis of failures of rolling element bearings is very important for improving both the reliability and safety of the rotating machinery that is widely used in the industry
We present a feature extraction method based on the Shannon entropy, which is computed by combining stationary wavelet transform (SWT) and singular value decomposition (SVD)
Digital data is produced at 12,000 samples per second for normal bearing (NB) samples and failure samples: inner race fault (IRF), outer race fault (ORF), and ball fault (BF)
Summary
Diagnosis of failures of rolling element bearings is very important for improving both the reliability and safety of the rotating machinery that is widely used in the industry. In order to achieve early diagnosis, we need to identify those hidden patterns that provide us with high-quality information regarding the bearing fault features. Extracting those features from non-stationary and non-linear vibration signals under time-varying speed conditions is not an easy task, and commonly used techniques for feature extraction are not accurate enough. We can find empirical mode decomposition (EMD) [1] and wavelet transform (WT) [2,3]. The EMD method can decompose a signal into a sum of intrinsic mode functions (IMFs) according to the oscillatory nature of the signal [4]. From signal decomposition methods, such as those above, different features can be calculated, such as energy entropy [9], permutation entropy [10], kurtosis value [11,12], relative energy [13], Entropy 2017, 19, 541; doi:10.3390/e19100541 www.mdpi.com/journal/entropy
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
