基于因子函数的α-IHS抽样方法

核电厂概率安全分析中,由于数据源存在不确定性,导致无法进行准确评估,因此需开展不确定性分析。样本在空间分布的均匀特性是不确定性分析的关键因素,不同的样本分布导致不确定分析结果差异较大。传统的拉丁超立方抽样方法在样本空间分布均匀性方面未进行优化,改进分布式超立方抽样(IHS)方法通过保持样本点之间的最优距离来实现空间均匀分布,但其最优距离只能在理想分布下达到最优。为改进IHS设计上的缺陷,提出了基于因子函数的α-IHS方法,利用修正因子α来优化IHS方法中的最优距离。结果表明,该方法较IHS方法具有更均匀的空间分布,提高了抽样效率。

α-IHS Sampling Method Based on Factor Function

In the probabilistic safety assessment of nuclear power plant, as for the uncertainty of data source, it’s essential to launch uncertainty analysis. As a key factor of uncertainty analysis, the distribution of sample space would have a great effect on results of uncertainty analysis. The Latin hypercube sampling method doesn’t make any improvement in space-filling of sample space, the improved distributed hypercube sampling method has improved it by keeping the optimal distance of sample points, but the optimal distance could only be achieved in ideal situation. In order to solve such a problem, α-IHS method based on factor function was put forward to optimize the optimal distance. The results prove that the α-IHS method has advantages over IHS method in a better distribution, and improves sampling efficiency.