基于Bayes混合验前分布的成败型产品可靠性评估

针对小子样成败型产品可靠性评估问题,引入混合验前分布构造方法,推导了Bayes混合验前分布下成败型产品可靠度参数的验后概率分布函数,研究了验前分布参数对于验后估计的影响,并总结了验前分布参数选择的一般性原则。分析对比了传统Bayes方法和混合验前分布方法的参数验后估计,其结果表明后者能够有效避免验前信息淹没现场信息的问题。对验后均方误差的分析表明混合验前分布方法能够一定程度上改善估计的效果。与幂验前方法的对比结果表明,当验前样本容量较大时,混合验前分布方法的估计效果优于幂验前方法。

Reliability Assessment for Device with Only Safe-or-failure Pattern Based on Bayesian Hyprid Prior Approach

For the reliability evaluation of device with only safe-or-failure pattern, a hyprid prior approach is introduced, and the posteriori probability distribution function of reliability parameter is deduced by Bayesian method. The influence of prior distribution parameters on the posterior estimation is considered, and the general rules of choosing these parameters are summarized. The parameter posterior estimates of traditional Bayesian method and hyprid prior approach are compared. The results indicate that the hyprid prior approach can avoid the problem of that the prior data inundate the posteriori data effectively. The analysis of posteriori mean square error demonstrates that the hyprid prior approach can improve the performance of estimation to some extent. Through the comparison with power prior approach, it is proved that the estimation effect of hyprid prior approach is better than that of power prior approach, especially when prior sample size is larger.