论文部分内容阅读
针对序贯分支仿真筛选实验中静态效应验证方法(V-SE)实验效率低且在仿真因子较多时实验效力难以保证的两方面不足,提出了基于动态效应(V-DE)和基于不重要因子(V-UF)的两种验证方法,V-DE在元模型估计中采用饱和设计并在配对检验中使用动态效应检验统计量,VUF仅对不重要因子的两组极端水平进行实验而不需估计元模型.蒙特卡罗仿真的三种方法对比实验以及基于实际背景仿真案例的应用表明,V-DE和V-UF在获得理想实验效力的同时极大提高了仿真实验的效率,大大降低了仿真工作者的实验成本,具有良好的实际应用价值.
In order to overcome the shortcomings of the V-SE experiment, which is inefficient and difficult to guarantee the experimental effectiveness when there are many simulation factors, a new algorithm based on dynamic effect (V-DE) and based on unimportant factors (V-UF), V-DE uses a saturated design in metamodel estimation and a dynamic effect test statistic in paired tests, while VUF tests only two extreme levels of unimportant factors without Estimation of Meta-Model Monte Carlo simulation of the three methods of comparison experiments and the application based on the actual background simulation case shows that, V-DE and V-UF achieve the desired experimental effectiveness while greatly improving the efficiency of simulation experiments, greatly reducing the Experimental cost of simulation workers, with good practical value.