Improved inverse Gaussian process and bootstrap: Degradation and reliability metrics. (October 2018)

• Record Type: Journal Article
• Authors: Guo, Jingbo; Wang, Changxi; Cabrera, Javier; Elsayed, Elsayed A.
• Is Part Of: Reliability engineering & system safety. Volume 178(2018)
• Page Start: 269
• Page End: 277
• Publication Date: 2018-10
• Subjects: Improved inverse Gaussian process -- Degradation -- Reliability -- Bootstrap -- Remaining life; 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.2018.06.013
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: The IIG model captures dependency of degradation increment on degradation state effectively. The IIG model predicts failure probability and residual life accurately. A new degradation model based on bootstrap for non-parametric data. The IIG and bootstrap models are validated with crack growth data and simulated data. Abstract: The inverse Gaussian (IG) process is commonly used in modeling monotonically increasing degradation processes. Traditional degradation modeling considers the process parameters as functions of time and environmental conditions. However, in many practical situations, the degradation increment in the next time interval may depend on degradation state at the current time interval. Therefore, in this paper, we propose an improved inverse Gaussian (IIG) process which considers the dependency between degradation increments and prior degradation states. Reliability metrics of the IIG process are estimated and validated using crack length growth data as well as simulated degradation data. Results show that the proposed model provides more accurate reliability metrics than the IG process model. Bootstrap of degradation increments or detrended degradation increments is introduced as a supplementary method to estimate the remaining life probability interval.