Adaptive Bayesian prediction of reliability based on degradation process. Issue 8 (3rd August 2022)

• Record Type: Journal Article
• Authors: Wang, Jun; Wang, Dianpeng; Tian, Yubin
• Is Part Of: Communications in statistics. Volume 51:Issue 8(2022)
• Page Start: 4788
• Page End: 4798
• Publication Date: 2022-08-03
• Subjects: Adaptive prediction -- Bayesian conditional c-optimal -- Bayesian reliability estimation -- Degradation process -- Multivariate normal-gamma prior; 62F15; Mathematical statistics -- Periodicals; Mathematical statistics -- Data processing -- Periodicals; Digital computer simulation -- Periodicals; 519.5
• DOI: 10.1080/03610918.2020.1749661
• Languages: English
• ISSNs: 0361-0918

Abstract

Abstract: For long-time running electric devices used in satellites, the accurate reliability prediction is crucial in engineering. The reliability of these devices is often directly related to the degradation of a performance characteristic. However, the problem about predicting the reliability of these devices based on a subset which is chosen from the real-time data flow adaptively has received scant attention in academic research. In this paper, an adaptive Bayesian conditional c-optimal criterion is proposed to select observations from the real-time data flow effectively. The conjugate prior which is described as MNG for the parameters in the model is derived. Then, based on the Bayesian conditional c-optimal criterion and the MNG conjugate prior, an approach to choose a subset of data, which makes the prediction robust, is suggested. Based on the simulated data from emulator created by Beijing Spacecrafts, an illustration and some simulations are done to study the performance of the proposed method for predicting the reliability of the devices from 16 to 20 years. The results show that our proposed method with MNG conjugate prior performs better than the local c-optimal method and the Bayesian method with Jeffreys's non-informative prior.