One-shot device test data analysis using non-parametric and semi-parametric inferential methods and applications. (May 2022)

• Record Type: Journal Article
• Authors: Zhu, Xiaojun; Balakrishnan, N.
• Is Part Of: Reliability engineering & system safety. Volume 221(2022)
• Publication Date: 2022-05
• Subjects: Accelerated life-test -- Expectation–Maximization algorithm -- Kaplan–Meier estimator -- Nelson–Aalen estimator -- Non-parametric inference -- One-shot device test -- Proportional hazards model -- Semi-parametric inference; 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.2022.108319
• Languages: English
• ISSNs: 0951-8320

Abstract

Abstract: A one-shot device, such as an automobile airbag, electro-explosive unit or munition, is a product that can be used only once. Its actual lifetime is unobservable, rendering the corresponding reliability analysis quite challenging. In this paper, two non-parametric methodologies—maximum likelihood estimation via EM-algorithm and Nelson–Aalen based estimation are developed for identical testing environment on one-shot devices. The EM-algorithm is usually used for unobserved failure times under some specific parametric models. But, here the EM-algorithm is adopted for the number of failures in each time interval in a non-parametric manner. Next, a semi-parametric approach, based on proportional hazards assumption, is developed for nonidentical testing environments, such as under an accelerated life-test. A Monte Carlo simulation study is then carried out for evaluating the performance of the inferential methods developed here. Finally, two data sets are analyzed for illustrative purpose. Highlights: Reliability analysis for one-shot devices data. Non-parametric and semi-parametric models are more robust. Expectation–Maximization algorithm with iteration having closed expression is more efficient. Statistical inference for the associated models.