Fourth-order kernel method for simple linear degradation model. Issue 1 (2nd January 2018)

• Record Type: Journal Article
• Authors: Arif, Osama H; Eidous, Omar
• Is Part Of: Communications in statistics. Volume 47:Issue 1(2018)
• Page Start: 16
• Page End: 29
• Publication Date: 2018-01-02
• Subjects: Classical kernel method -- Degradation -- Failure time -- Fourth-order kernel method -- Maximum likelihood -- Ordinary least square; 62N02 -- 62G07; Mathematical statistics -- Periodicals; Mathematical statistics -- Data processing -- Periodicals; Digital computer simulation -- Periodicals; 519.5
• DOI: 10.1080/03610918.2016.1186183
• Languages: English
• ISSNs: 0361-0918

Abstract

ABSTRACT: Degradation analysis is a useful technique when life tests result in few or even no failures. The degradation measurements are recorded over time and the estimation of time-to-failure distribution plays a vital role in degradation analysis. The parametric method to estimate the time-to-failure distribution assumed a specific parametric model with known shape for the random effects parameter. To avoid any assumption about the model shape, a nonparametric method can be used. In this paper, we suggest to use the nonparametric fourth-order kernel method to estimate the time-to-failure distribution and its percentiles for the simple linear degradation model. The performances of the proposed method are investigated and compared with the classical kernel; maximum likelihood and ordinary least squares methods via simulation technique. The numerical results show the good performance of the fourth-order kernel method and demonstrate its superiority over the parametric method when there is no information about the shape of the random effect parameter distribution.