估计感度分布中位数的两阶段优化序贯试验设计方法研究  

Two-Phase Sequential Optimal Design for Estimating the Median of Sensitivity Distribution

作　　者：白天 侯旭 王典朋 BAI Tian;HOU Xu;WANG Dianpeng(School of Mathematics and Statistics,Beijing Institute of Technology,Beijing 100081)

机构地区：[1]北京理工大学数学与统计学院,北京100081

出　　处：《系统科学与数学》2020年第2期341-354,共14页Journal of Systems Science and Mathematical Sciences

基　　金：青年科学基金项目(11801034)资助课题。

摘　　要：感度试验在爆炸科学领域有非常广泛的应用,其中位数是含能材料的一个关键性能指标.高效估计感度分布中位数是实际工程应用中的重要问题.文章提出一种两阶段的优化序贯试验设计方法来估计感度分布的中位数,并将正交设计与两阶段优化序贯试验设计方法相结合给出了筛选影响含能材料安定性重要因子的试验设计方法.广泛的模拟比较验证了所提方法在小样本情况下可以获得感度分布中位数的准确估计,并且对模型假设和初始参数猜测具有一定的稳健性.当因子效应不是太小的时候,因子筛选方法可以准确地筛选出有显著影响的因子.Sensitivity tests are widely used in explosion science and the median is a key indicator of energetic materials.Because of its versatile applications,an efficient procedure needs to be proposed.In the paper,we propose a two-phase sequential optimal design to estimate the median of the sensitivity distribution.Furthermore,we combine our design with orthogonal design and propose a method to select significant factors.Extensive simulation comparisons show that our method outperforms in estimation of the median and in terms of the robustness of model and initial parameter assumptions.When the factor effect is not too small,our method could select significant factors effectively.

关 键 词：感度试验 两阶段优化序贯试验 感度分布中位数 

分 类 号：O212.6[理学—概率论与数理统计]