A hybrid prognostic method for system degradation based on particle filter and relevance vector machine. (June 2019)

• Record Type: Journal Article
• Authors: Chang, Yang; Fang, Huajing
• Is Part Of: Reliability engineering & system safety. Volume 186(2019)
• Page Start: 51
• Page End: 63
• Publication Date: 2019-06
• Subjects: Prognostics -- Particle filter -- Relevance vector machine -- Deterministic prediction -- Prediction interval -- Lithium-ion battery; 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.2019.02.011
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: The proposed prognostic method can provide accurate and stable RUL prediction. The proposed prognostic method can construct a prediction interval to assess the prediction uncertainty. Four types of comparative experiments are performed to verify the wide applicability of the proposed method; Reliable prognostic result can be provided by proposed method to ensure the system reliability. Abstract: Prognostics of the remaining useful life has become a critical technique to ensure the reliability and safety of system, however, due to the uncertainty of system degradation, the prognostic result is usually not so satisfactory. To solve this problem, a hybrid prognostic scheme with the capability of uncertainty assessment is proposed in this paper, which combines particle filter (PF) and relevance vector machine (RVM). The prognostic result comprises a set of deterministic prediction values to represent the degradation process and a prediction interval to evaluate the prediction uncertainty. In order to examine the performance of the proposed hybrid method, four types of comparative experiments based on two types of lithium-ion battery datasets and two degradation models are performed. The experimental results show that the proposed hybrid scheme is a reliable prognostic method which can ensure the accuracy of the deterministic prediction result and provide precise assessment for the prediction uncertainty.