A MULTIRUN STEP-STRESS MODEL FOR TREND RENEWAL DATA WITH APPLICATIONS TO LIFETIME ASSESSMENT FOR RECHARGEABLE BATTERIES

• 成果类型: Article
• 署名作者: Tseng, Sheng-Tsaing; Hsu, Nan-Jung; Wu, Chien-Chi
• 署名单位: National Tsing Hua University
• 刊物名称: ANNALS OF APPLIED STATISTICS
• ISSN/ISSBN: 1932-6157
• DOI: 10.1214/24-AOAS1993
• 发表日期: 2025
• 页码: 1603-1622
• 关键词: accelerated degradation tests; prediction; DESIGN

摘要

Conducting a cost-efficient lifetime-testing plan to timely assess lifetime information of a highly reliable product is often a challenging task in the manufacturing industry. Motivated by a case study of rechargeable lithium-ion batteries, this paper introduces a multirun k-level step-stress experiment, running under different stresses in repeated cycles, to collect and analyze the degradation data of reusable highly-reliable products. Specifically, we formulate the battery capacity over recharge cycles as a counting process and adopt a trend renewal process (TRP) to characterize the degradation patterns of capacity varying with the stress level of the accelerated factor. By using a Markovian property on cumulative exposure in our counting process, the degradation data observed in a multirun k-level step-stress TRP model can be converted equivalently to corresponding k constant-stress TRP models. This connection allows us to estimate the parameters using maximum likelihood and to infer with uncertainty quantification the end-of-performance (EOP) of batteries at normal-use conditions. This novel method is shown to be efficient for the lifetime assessment of reusable products.