应用分层模型进行核电站设备可靠性参数估计

Calculation of Reliability Parameters of Nuclear Power Plant with Hierarchical Model

目前国内对于从多个核电站统计的设备失效数据进行各特定核电站设备可靠性参数估计的方法研究尚少。本文研究了用于可靠性参数估计的分层模型以及实现分层模型的两种方法:带Kass-Steffey修正的参数经验贝叶斯方法和马氏链蒙特卡洛方法。以设备需求失效的稀少失效数据样本为例,推导了带Kass-Steffey修正的Beta-Binomial模型原理并编程求解,研究了马氏链蒙特卡罗方法及软件计算,对比了核电站后验失效概率的计算结果。计算表明:两种方法得到的部分失效概率后验估计的均值相差0~25%;95分位值相差5%~15%,两种方法都可用于稀少数据的样本估计。

There is little research on the calculation methods of evaluation reliability parameters directly from our own statistics failure data.The parametric empirical Bayes model with Kass-Steffey adjustment and the MCMC method are two Bayesian methods to estimate the reliability parameters.Take the failures on demand with sparse data as an example,this paper derives the principle of Beta-Binomial with Kass-Steffey adjustment,adopts two methods to calculate the hyperparameter and plant-specific posterior parameter.It shows that posterior mean of NPPs change 0-25%,posterior95 th change 5% -15%.Two methods are all suitable for estimating the reliability parameters from sparse failure data.