Academy of Mathematics and Systems Science, CAS
Colloquia & Seminars

Speaker:	王成杰博士，同济大学
Inviter:
Title:	Recoverability effects on reliability assessment for accelerated degradation testing
Language:	Chinese
Time & Venue:	2022.01.13 14:00-15:00 腾讯会议:706-785-429
Abstract:	Accelerated Degradation Testing (ADT) provides an efficient experimental approach to collect lifetime-related data for the reliability assessment of highly reliable products under normal use stress. Recoverability, which occurs in many typical failure modes when online in-chamber measurements have to be replaced by offline ex-chamber measurements with stress released, is an important factor that may affect the accuracy of reliability estimation from ADT data. Nonetheless, the presence of recoverability has not been adequately considered in traditional methods, leading to an over-optimistic estimation of lifetime and reliability. The linkage between recoverability and such inferential bias has not been theoretically studied systematically. In this study, recoverability is explicitly incorporated into the ADT modeling framework. Without loss of generality, the Wiener process is adopted as the basis for the proposed degradation model, superimposed with a cumulative recovery. Theoretical results show that the Mean Time To Failure (MTTF) will be overestimated with recoverability neglected, which leads to poor lifetime-centered decision-making. For finite and even small sample sizes, this conclusion is not intuitively certain but the chance is still high and the corresponding overestimation probability can be calculated explicitly. All theoretical conclusions are validated by simulation studies, in which the MTTFs are overestimated from 6% to 42% under different parameter settings. The real-world application of semiconductor products shows that even slight recoverability could lead to an obvious overestimation of MTTF and the proposed model provides a convenient way to derive more accurate assessment results. 报告人简介：Chengjie Wang received the BS degree in mathematics & applied mathematics from Nanjing University in 2016 and a PhD in statistics from the Academy of Mathematics and Systems Science (Chinese Academy of Sciences) in 2021. His research interests include statistical modeling and analysis for complex systems, machine learning, and reliability. He is now a postdoctoral researcher in the Tongji University.