基于正交试验的元模型输入参数筛选方法

Study of Input Parameters Filtration Method for Simulation Metamodel Based on Orthogonal Experiment

输入参数筛选工作可有效降低元模型的构建维度。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,存在的难以获取原始模型解析模型和难以反映多输入参数综合影响的两个突出问题,为准确筛选出对元模型输出参数影响显著的输入参数,提出了基于正交试验设计的元模型输入参数筛选方法,该方法结合正交表构造方法,综合运用极差分析、方差分析、输出参数变化趋势三种方法综合研究原始模型输入参数对输出参数的影响程度。最后以螺线管制动器激励电流元模型的构建为案例,介绍了输入参数筛选方法的应用过程。结果表明,该方法可有效减少元模型输入参数的数量。

Input parameters filtrating will reduce the difficulty effectively for metamodel construction. The problems when traditional sensitivity analysis method was applied to input parameters filtration of metamodel, such as the getting of analytic model, and the general affection of many parameters, were analyzed. In order to filter out the input parameters that did not affect the output parameters for metamodel evidently, the input parameters filtration method based on orthogonal experiment was proposed which synthetically explored the affection of input parameters for original model output parameter through utmost difference, square difference and output parameter changing trend based on orthogonal table construction. Aiming at the metamodel building of solenoid arrester stimulant current simulation, the input parameters filtrating process was introduced in detail. The result validates that the method can reduce the difficulty of metamodel building effectively.