A predictive sliding local outlier correction method with adaptive state change rate determining for bearing remaining useful life estimation. (September 2022)

• Record Type: Journal Article
• Authors: Wang, Han; Wang, Dongdong; Liu, Haoxiang; Tang, Gang
• Is Part Of: Reliability engineering & system safety. Volume 225(2022)
• Publication Date: 2022-09
• Subjects: Local outlier correction -- Remaining useful life prediction -- First predicting time -- Exponential model; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2022.108601
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: A predictable sliding local outlier detection method for time series data. Propose an algorithm to determine the optimal FPT based on degenerate state changes. The exponential degradation model is optimized to improve the prediction accuracy. The superiority of the proposed method is demonstrated based on the PRONOSTIA dataset. Abstract: The accurate estimation of remaining useful life (RUL) is significant for the operation, maintenance, and avoidance of unplanned downtime of rotating machinery. To improve the prediction accuracy of RUL, this paper proposes a predictive sliding local outlier correction (PSLOC) method with adaptive state change rate (SCR) determining algorithm for bearing RUL prediction. The proposed method addresses some key issues in existing bearing health prognosis methods that affect the prognostic performance. First, the PSLOC method is proposed, which can eliminate the random spurious fluctuations in the degradation process. It not only ensures the integrity of the degradation trend, but also avoids the influence of abnormal fluctuations in the time-series data on the judgment of the first predicting time (FPT) and the RUL prediction accuracy. Second, the proposed SCR algorithm can be actively adjusted according to the changes in the degradation process to achieve an accurate FPT determination. Finally, the simulated degradation algorithm is used to improve the exponential model to reduce the random errors in the degradation process and improve the prediction accuracy. The PRONOSTIA dataset is used as a case study to illustrate the superiority and effectiveness of the proposed method.