CAJ | 학술논문

为了提高剩余寿命预测的可信度，针对进行过加速老化试验的产品，提出利用Gamma过程参数的非共轭先验分布进行Bayesian统计推断的剩余寿命预测方法。将加速老化数据作为先验信息，利用Gamma过程进行老化建模，通过加速因子获得形状参数在工作应力下的折算值，使用 Anderson‐Darling 统计量确定随机参数的先验分布。将产品工作中的少量实测数据作为现场信息，利用基于Gibbs抽样的Markov Chain Monte Carlo（MCMC）仿真得到参数的后验均值。以某型导弹电连接器为例说明了该方法的研究意义和工程应用价值。
위료제고잉여수명예측적가신도，침대진행과가속노화시험적산품，제출이용Gamma과정삼수적비공액선험분포진행Bayesian통계추단적잉여수명예측방법。장가속노화수거작위선험신식，이용Gamma과정진행노화건모，통과가속인자획득형상삼수재공작응력하적절산치，사용 Anderson‐Darling 통계량학정수궤삼수적선험분포。장산품공작중적소량실측수거작위현장신식，이용기우Gibbs추양적Markov Chain Monte Carlo（MCMC）방진득도삼수적후험균치。이모형도탄전련접기위례설명료해방법적연구의의화공정응용개치。
A method based on Bayesian inference with non‐conjugate prior distributions of parameters of Gamma processes was proposed for the products which had gone through accelerated aging tests in order to improve the reliability of remaining useful life prediction .The accelerated aging data was considered as prior information and Gamma processes were utilized to model the aging data .After the transformed values of parameters at the normal working stress were obtained by acceleration factors ,the Anderson‐Darling statistic was applied to determine the prior distributions of two parameters .Few testing data at working was considered as field information ,and the posterior means of parameters were obtained by Markov Chain Monte Carlo simulation based on Gibbs sampling .The research significance and engineering value were demonstrated through a missile electrical connector .