摘要
可靠性数据是核电站概率安全评价的基础,目前国内核行业通常使用经验贝叶斯方法估计可靠性参数。然而极大似然函数对于稀少样本的估计存在求解困难的问题。马氏链蒙特卡洛方法提供了另一种可用于可靠性参数估计的分层贝叶斯方法,其从后验分布大量抽样进而推断总体特征。采用基于马氏链蒙特卡洛方法的Win BUGS软件计算了核电站需求失效型稀少样本的超参数。针对68个核电站辅助给水电动部分启动需求失效数据,计算得到启动失效的超参数α为0.192 5,β为46.77,以及各个核电站辅助给水电动部分启动需求失效的后验失效概率。最后对Win BUGS软件中马氏链的收敛性进行讨论。
Reliability data is the foundation of probabilistic safety assessment in NPP. MCMC method is another hierarchical bayes which can avoid the complicated solutions for the parametric empircal bayes models. The WinBUGS based on MCMC was Opplied to calculate the hyperparameter of demand failure data. Based on the data of failure to start of AFW motor-driven segments at 68 plants,that the hyperparameter α is 0. 1925 and β is 46. 77 were found out,and the posterior failure probability of 68 plants were obtained. At last the convergence of markov chain was discussed. It hopes providing another way to calculate the reliability parameter of nuclear power plant in China.
出处
《科学技术与工程》
北大核心
2016年第1期151-154,共4页
Science Technology and Engineering
基金
国家科技重大专项(2013ZX06002001-008)资助
