Transient waveform matching based on ascending multi-wavelets for diagnostics and prognostics of bearing deterioration. (January 2022)

• Record Type: Journal Article
• Authors: Jiang, Kuosheng; Li, Zhixiong; Zhou, Yuanyuan; Sarkodie-Gyan, Thompson; Li, Weihua
• Is Part Of: ISA transactions. Volume 120(2022)
• Page Start: 330
• Page End: 341
• Publication Date: 2022-01
• Subjects: Transient waveform -- Bearing deterioration -- Multi-wavelet -- Ensemble learning; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2021.03.016
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: Run-to-failure experiment is efficient and effective to investigate bearing deterioration process. Periodic transient waveform carries rich information of health conditions of bearings but the transient waveform matching is a challenging problem for evaluating bearing fatigue life because the shapes and parameters of the waveform vary with the evolution of the bearing degradation. A wavelet function such as a Morlet wavelet is able to extract essential features from the transient waveform but limited to a single transient component. The multi-wavelet may provide a solution to fit a variety of primary components in the transient waveform, so as to track the degradation trend of the bearing; however, very limited work has been done to address this issue. To bridge the research gap in the transient waveform matching, a novel ascension multi-wavelet method is proposed in this paper for diagnosing the undergoing degradation state and predicting the remaining useful life (RUL) of the bearings. Firstly, the transient waveform was matched using the combination of multiple wavelets. Then, the entropy of the multiple-wavelet signal was calculated to quantify the periodic transients to generate a monotone trend of the bearing degradation. The degradation state of the bearing was identified using the entropy. Lastly, the ensemble learning method was employed to establish an RUL predictor. Both simulation and experiments were carried out to evaluate the proposed method. The analysis results demonstrate satisfactory diagnostics and prognostics performance of the proposed method. The RUL prediction accuracy of the multi-wavelet matching is better than that of the single-wavelet matching. Highlights: The transient waveform was matched using the combination of multiple wavelets. The degradation state of the bearing was identified using the entropy. The ensemble learning method was employed to establish an RUL predictor.