Abstract: In order to resolve the problems when traditional sensitivity analysis method is applied to input parameters filtration for metamodel, the input parameters filtration method is presented based on information entropy theory. The parameters system applied to metamodel input parameters filtration is constructed, including uncertainty ratio, uncertainty ratio variety, and uncertainty ratio cumulate variety; the analysis process for input parameters filtration is presented; the composable model constructed by several submodels, as well as it’s condition entropy, is researched through Bayesian network. The input parameters filtration method is applied to the simulation of equipment battlefield damage resulted in by a fighter plane attacking, including single model input parameters filtration process and composable model input parameters filtration process.

Key words: metamodel, simulation, entropy, uncertainty, composable model

摘要： 针对传统的灵敏度分析方法应用于元模型输入参数筛选时存在的突出问题，基于信息熵理论提出了元模型输入参数筛选方法。以条件熵为基础，提出了用于元模型输入参数筛选的参数体系，包括不确定度、不确定度浮动、不确定度累积浮动，以及元模型输入参数筛选分析过程；采用贝叶斯网络技术研究了多个子模型构成的组合模型描述问题，以及相应的条件熵计算问题；针对航空兵轰炸对于装备的损伤仿真问题，对单一模型和组合模型的输入参数筛选问题开展了案例研究。

关键词: 元模型, 仿真, 熵, 不确定性, 组合模型