A Bayesian approach to modeling two-phase degradation using change-point regression. (February 2015)

• Record Type: Journal Article
• Authors: Bae, Suk Joo; Yuan, Tao; Ning, Shuluo; Kuo, Way
• Is Part Of: Reliability engineering & system safety. Volume 134(2015:Feb.)
• Page Start: 66
• Page End: 74
• Publication Date: 2015-02
• Subjects: Degradation modeling -- Change-point regression -- Failure-time distribution -- Gibbs sampling -- Hierarchical Bayesian modeling; 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.2014.10.009
• Languages: English
• ISSNs: 0951-8320

Abstract

Abstract: Influenced by defects or contaminants remaining after a series of manufacturing processes, the degradation paths of some products exhibit two-phase patterns over the testing period. This paper proposes a hierarchical Bayesian change-point regression model to fit the two-phase degradation patterns, and derives the failure-time distribution of a unit that is randomly selected from its population. A Gibbs sampling algorithm is developed for the inference of the parameters in the change-point degradation model, as well as for the prediction of the failure-time distribution of the randomly selected unit. The proposed approach is applied to the degradation paths of plasma display panels (PDPs) presenting the two-phase pattern. Abstract : Highlights: We develop hierarchical Bayesian models for two-phase degradation phenomena. We model the two-phase degradation paths by a change-point regression approach. Gibbs sampling algorithms are developed for parameters and lifetime estimation. We analyze the two-phase degradation data for plasma display panels. The proposed methodology is effective for analyzing two-phase degradation.