基于Jeffreys先验的参数相关退化可靠性评估

DEPENDENT PARAMETERS DEGRADATION RELIABILITY ASSESSMENT BASED ON JEFFREYS NONINFORMATIVE PRIOR PARAMETERS

通过Jeffreys无信息先验分布描述了Gamma退化过程中参数的相关性,由贝叶斯模型得到各参数满条件分布,使用马尔科夫链蒙特卡洛(Markov Chain Monte Carlo,MCMC)方法得到参数后验期望估计,最后给出可靠度评价模型。工程实例表明,所得可靠性评估较独立情形更为保守,能够更早地给出产品修理建议。同时,仿真表明,可靠度要求越高,相关与独立情形寿命估计结果偏差越大,0.9999可靠度下偏差率最大可达9.26%。

The correlation between the parameters of the Gamma degradation process was described by the Jeffreys uninformative prior distribution.And the Bayesian model was used to obtain the full conditional distribution of each parameter.The MCMC method was used to get parameter posterior expectation estimates.Finally,reliability was calculated according to the engineering examples and 100 simulations,the obtained reliability assessment was more conservative than the independent case in engineering practice.Thus,the product repair suggestion could be given earlier.And the higher the reliability requirement was,the greater the deviation between the estimation results of the correlated case and the independent case was,and the life estimation error rate under the reliability of 0.9999 was up to 9.26%.