摘要
核动力个别设备(如电动泵)在其运行历史上无失效记录或仅有很少的失效记录,仅用小样本数据反映设备可靠性参数的总体分布有所欠缺。本文提出"两步走"方法,即对核动力设备数据进行处理时参考同堆型核电站的通用数据库,将属于无信息先验的核动力设备数据处理转化为有信息先验的处理。应用Jeffreys先验模型对核动力电动泵进行贝叶斯推断,通过图检验评价模型复现观察数据的能力,结果表明模型可以完全复现观察数据具有良好的预计能力,经分析建议将第4组数据剔除后再对电动泵失效数据进行贝叶斯推断。
There is no or few failure records about some individual device（such as motordriven pumps） of one type of nuclear power equipment.It was not appropriate to use rare sample to reflect the population distribution of the device reliability parameters in PSA.The paper presented ＂two steps＂ method when dealing with the data of nuclear device the general data based on the same type nuclear power plant is used.And this could translate non-informative prior into informative prior.The motor-driven pumps of nuclear power plant is taken as Jeffreys prior information as an example and inspection the model ability of replicating failure data by graph is analyzed.Results show that Jeffreys prior model could replicate all the records of motor-driven pumps operation failures and had an excellent predicted ability.
出处
《核科学与工程》
CAS
CSCD
北大核心
2016年第3期419-423,共5页
Nuclear Science and Engineering
基金
核反应堆系统设计技术国家重点实验室基金资助项目(HT-JXYY-02-2014002)
作者简介
郭海宽（1988-），男，河北邢台人，博士研究生，现从事核科学与技术方面研究
