基于Dirichlet先验分布的加速寿命试验的Bayes分析  

Bayesian Analysis of Dirichlet Prior Distribution-Based Accelerated Life Test

作　　者：谭源源[1] 张春华 陈循[1] 

机构地区：[1]国防科技大学机电工程与自动化学院机电工程研究所,长沙410073

出　　处：《宇航学报》2010年第6期1678-1684,共7页Journal of Astronautics

基　　金：国家部委资助项目(203020102)

摘　　要：竞争失效场合加速寿命试验统计分析问题正在成为研究的热点,但目前的方法主要采用极大似然估计(MLE),在小样本情形下统计精度不高。针对这一问题,以Dirichlet分布综合产品的各类可靠性先验信息,并充分考虑了试验数据的非完整性,建立了基于Dirichlet先验分布的Bayes分析方法。最后通过MonteCarlo仿真和应用算例验证方法的有效性。结果表明,由于综合了有用的先验信息,从而扩大了统计信息量,因此在小样本情形下本文方法比极大似然估计方法具有更加优良的估计性质。In the presence of competing failure modes an accelerated life test （ALT） has been taken into account for some time.However,most of the methods proposed in relevant work are mainly based on MLE （Maximum Likelihood Estimations）,which is not very appropriate in the case of small sample size.Aiming at this problem,a Dirichlet prior distribution based Bayes method is proposed.The method integrates prior information of relevant reliabilities with Dirichlet distribution,and fully considers the incompleteness of test data.Furthermore,Monte Carlo simulation and case study are performed to check validity of the method,and results show that the method provides better estimators than MLE by incorporating available prior information when sample size is small.

关 键 词：加速寿命试验 竞争失效 Dirichlet分布 BAYES分析 MONTE Carlo仿真 

分 类 号：TB114.3[理学—概率论与数理统计]