Analysis of Common Cause Failure Data in Nuclear Power Plant Safety Assessment

核电厂安全评估中的共因故障数据分析

• 批准号: 9700527
• 负责人: Paul Kvam
• 金额: $ 11.49万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1997
• 资助国家: 美国
• 起止时间: 1997-09-15 至 2000-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9700527&HistoricalAwards=false
• 关键词: Analysis; Common; Cause; Failure; Data

This grant supports the development of new statistical methods for analyzing common cause failure data for complex systems, such as nuclear power plant component groups. For the purpose of modeling common cause failures, the research can lead to the development of a binomial failure rate mixture model, based on a parametric shock model developed for power plant risk assessment. Beyond this mixture model estimator, various results will be pursued, including both non-parametric and parametric estimators; confidence bands and other measures of uncertainty; general solutions for systems of non identical components; and, pertinent limit properties. The mixture model will serve as a point of departure to further research into statistical analyses based on dependent failure models. This research also investigates techniques of mapping that replace missing data with estimated data in the risk assessment, and will help to characterize the uncertainty involved with various models used for the mapping procedure using statistical theory in addition to Monte Carlo simulation. The original binomial failure rate model has been criticized due to the many existing databases for which the model fit is poor. The realistic generalization in this investigation is more complex than the original model or its non-parametric alternatives, but model fit can be greatly improved. There exists a strong foundation of qualitative analysis in the engineering literature (screening analysis, fault tree analysis, etc.), but little guidance with respect to some quantitative aspects common cause failure research. If successful, this research will contribute to quantitative risk analyses in the nuclear safety field.

这笔赠款支持开发新的统计方法，用于分析复杂系统(如核电站组件组)的常见原因故障数据。为了对常见原因故障进行建模，该研究可以在为电厂风险评估开发的参数冲击模型的基础上，开发二项故障率混合模型。除了这种混合模型估计器，还将寻求各种结果，包括非参数和参数估计器；置信带和其他不确定性度量；非相同组件系统的一般解；以及相关的极限性质。混合模型将作为进一步研究基于相依失效模型的统计分析的出发点。这项研究还研究了在风险评估中用估计数据替换缺失数据的映射技术，并将有助于利用统计理论和蒙特卡洛模拟来表征映射过程中使用的各种模型所涉及的不确定性。原有的二项式失效率模型由于现有的许多数据库模型对模型的拟合度较差而受到了批评。与原始模型或其非参数方案相比，本研究中的现实泛化更为复杂，但模型的拟合度可以大大提高。在工程文献中有很强的定性分析基础(筛选分析、故障树分析等)，但对于一些定量方面的共性故障研究却缺乏指导。如果成功，这项研究将有助于核安全领域的定量风险分析。