基于仿真的虚拟设计中的统计回归与优化研究  

Statistics Regression and Optimization in Virtual Design Based on Simulation

作　　者：潘军[1] 马登哲[1] 范秀敏[1] 金烨[1] 

机构地区：[1]上海交通大学机械工程学院CIM研究所,上海200030

出　　处：《计算机集成制造系统-CIMS》2004年第4期404-409,共6页

基　　金：国家自然科学基金资助项目(59889505)。~~

摘　　要：将统计回归技术和试验设计作为一种方法学引入虚拟设计过程,系统地总结了试验设计策略和统计回归模型的选择原则;基于复合材料结构数值分析过程,从模型精度、模型优化效率及模型透明度对试验设计策略和统计回归模型进行了研究和对比;最后,提出了全局近似和优化精度面向不同目标的启发式采样策略的改进方法。初步实践表明,该方法可优化基于仿真的虚拟设计,揭示了仿真规律,实现了仿真和优化算法的快速集成。The statistics regression and design of experiment technologies are introduced as a methodology into the virtual design process. The selection principle of design of experiment and statistics regression model is systematically concluded. Based on a numerical analysis process of composite material structure, strategies of design of experiment and regression model are discussed from model precision, optimization efficiency and model insight. Finally, a heuristic sampling method oriented to different targets is pointed to increase the model's global proximate and optimization precision. The primary practice showed that this method is practical for virtual design and optimization based on simulation, and realized rapid integration between optimization algorithm and simulation.

关 键 词：虚拟设计 统计回归 优化 复合材料结构 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]