基于复杂昂贵仿真的体系效能多目标优化

效能优化是实现体系结构设计、多方案配置等工作的重要途径.体系仿真系统通常具有组成结构复杂、连续离散混合、输入输出变量多、运行开销大等特点,导致效能优化面临多目标、混合变量、多峰值、低效率等问题,提出一种基于复杂昂贵仿真的体系效能优化算法.针对昂贵仿真问题,提出基于聚类与空间填充准则相结合的开发-探索序贯元模型策略;引入混合编码策略和相应遗传算子改进传统的NSGA-II (nondominated sorting genetic algorithm II)算法,提高多目标优化效率.与目前较流行的3种昂贵优化算法的对比分析表明,所提算法得到的优化结果具有较好的收敛性、多样性及分布性.最后,以某防空体系的效能仿真优化为例,验证所提方法的可行性和有效性.

Complex and expensive simulation based multi-objective optimization to system-of-system effectiveness

Effectiveness optimization is the important approach to design system-of-system structure and configure scenario schemes. The system-of-system simulation system is usually characterized by complex structure, continuous and discrete mixing, multiple input and output variables, and high operation cost, which leads to some problems of multi-objective, mixed variables, multi-peak and low efficiency, etc. existed in effectiveness optimization. A complex and expensive simulation-based approach to system-of-system effectiveness optimization is proposed. The exploitation- exploration based sequential metamodel strategy is presented coupling clustering with space filling criterion for expensive simulation. The hybrid coding strategy and the corresponding genetic operators are introduced to improve the multiobjective optimization efficiency of the traditional NSGA-II algorithm.The comparative analysis with the state-of-art compensive optimization algorithms shows that optimization results of the proposed algorithm have better convergence, diversity and distribution. Futhermore, the effectiveness simulation optimization for some air defense system-of-system is taken as an example to verify the feasibility and effectiveness of the proposed method.