【作者】 刘云飞；

【导师】 邱晓刚；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2009， 硕士

【摘要】 信息时代的军事变革引起了军事认知方式的深刻变化,军事科学规律的研究不再局限于对过去战争经验的总结和历史数据的归纳,通过作战仿真实验的方法可为信息化战争提供实时决策支持。随着作战仿真系统规模的日趋庞大,对仿真结果有显著影响的实验因素也越来越多,要用有限的实验次数获得尽可能多的信息就需要有科学的实验设计方法的指导。采用合适的实验设计方法并提供有效的实验设计工具,可以有效地减少仿真实验的次数、提高效率,并使得对仿真结果的分析有章可循。论文对正交设计等传统实验设计方法的原理进行研究,分析其在作战仿真实验中的应用与传统应用的异同,并提出了一些改进策略,使这些实验设计方法能够更有效地应用于作战仿真实验。为了评估正交设计等三种实验设计方法在作战仿真中的应用效果,论文采用EINSTein系统作为作战仿真环境以及数据来源。选择部分因子设计、正交设计、均匀设计这三种实验设计方法来安排仿真实验,作为获取训练样本的手段,并选用多项式回归建模以及神经网络建模这两种建模方法来建立EINSTein系统的元模型,预测其输入输出特性。通过对实验结果的综合分析,得出效果最佳的实验设计方法和建模方法的搭配,并分析各实验设计方法适用于何种仿真系统,对实验设计在作战仿真中的应用具有指导意义。最后,根据作战仿真实验管理系统实现多方案连续仿真的需求设计和实现了实验设计工具SIMDOE。实现了实验设计的智能算法,其能够根据仿真的需求选择合适的实验设计方法并直接给出实验安排。实验管理系统照此对实验条件进行修改,生成一组仿真方案并连续执行,实现了多方案连续仿真的智能化。更多还原

【Abstract】 The transformation of Military affairs in the information age arouses change of methods of cognition about Military affairs. Research of the rule of Military affairs will no longer be restricted to the summary of the experience of war and the conclusion of data in history, but focus on cognition and practice about future war. The number of remarkable factors increases greatly with the growth of the scale of simulation system. Therefore, some suitable experiment design methods should be introduced into the warfare simulation experiments in order to decrease the simulation time and improve the efficiency. These design methods also provide effective analytic technique to the simulation output.Some traditional experiment design methods such as Orthogonal Design are researched on this paper, meanwhile some improved tactics are put forward in order to make these methods more suitable and effective to the warfare simulation experiments.In order to evaluate the applicability of three typical design methods used in warfare simulation, Einstein System is introduced as the simulation environment and data source. Fraction Factorial Design, Orthogonal Design and Uniform Design are used to arrange the simulation experiments, and Regression Modeling and Neural Networks Modeling methods are used to build up the meta-model of the simulation system. A serial of experiments are done to evaluate the effect with different combination of design methods and modeling methods. The conclusion of the experiments could be heuristic for the application of Experiment Design in warfare simulation.At last, a tool (SIMDOE) is developed to fulfill the continuous-running requirement of the multi-project simulation. It could produce the experiment plan directly as long as the basic information is provided. The experiment management system could run the simulation continuously according to this plan.更多还原