基于MCMC方法的继电器加速寿命试验分析

Analysis for Accelerated Life Test of Relays Based on MCMC Method

针对继电器贮存寿命的可靠性评估问题,根据生存分析理论,提出基于Box-Cox变换的步进加速寿命试验模型。考虑多个环境应力,结合累积失效模型,运用基于Gibbs抽样的MCMC方法,给出先验分布为多元分布情形下继电器加速寿命试验模型参数的贝叶斯估计,并利用BUGS软件包进行仿真分析。为了与传统方法进行比较,给出参数的极大似然估计,并对两种方法进行模拟分析。模拟结果表明,提出的加速寿命试验模型及模型参数估计方法可有效提高继电器寿命评估的精度,特别适合小子样继电器的可靠性评估。

A step-stress accelerated life test model for storage reliability evaluation of relays was proposed under the Box-Cox transformation based on survival analysis theory. A cumulative exposure model with muhi-stress variables was assumed. The parameters＇ Bayesian estimations of this model were developed using Markov Chain Monte Carlo （MCMC） based on Gibbs sampling with a multivariate prior distribution. A numerical example was presented to illustrate the approach using the BUGS package, and some numerical comparisons of the Maximum Likelihood （ML） method and Bayesian method were also carried out. The simulation results showed that the proposed method had offered a quite accurate inference compared with the ML method especially in the case of small sample size.