基于改进多层交叉熵方法的核动力装置非能动系统可靠性评估

Reliability Evaluation of Passive System for Nuclear Power Plant Based on Improved Multi-Level Cross Entropy Method

针对非能动系统可靠性评估问题中重要抽样分布构造困难以及实际工程应用中标准多层交叉熵方法存在的缺陷,进行了算法结构改进并引入均匀性更好的"Halton序列"抽样方法,提出一种改进多层交叉熵方法,并以某型核动力装置非能动余热排出试验系统为例,给出改进多层交叉熵方法的性能验证分析和实例分析。计算结果表明,改进方法失效率估计结果的相对误差分布更集中、变异系数分布相近,在节省运算量的情况下具有更好的估计精度和稳健性;改进方法不需要额外设置平滑参数、可根据系统特征和采样情况提前结束评估过程,在实际工程应用中具有更强的适用性。

In view of the difficulty in constructing the importance sampling density in the reliability evaluation of passive system and the defects of the standard multi-level cross entropy method in practical engineering application,an improved multi-level cross entropy method is proposed by improving the algorithm structure and introducing the Halton sequence sampling method with better uniformity.Taking an experimental facility for passive residual heat removal system of a certain marine nuclear power plant as an example,the performance verification analysis and example analysis of the IMCE method are implemented.The calculation results show that,the relative error distribution of the improved method is more convergent and the distribution of the coefficient of variation is similar,and the improved method is with better estimation accuracy and robustness with less computation.Moreover,the improved method is with stronger applicability in practical engineering application,because there is no need to set additional smoothing parameter,and the evaluation process can be completed earlier according to the system characteristics and sampling conditions.