摘要
基于图形变形法探索固体火箭发动机优化设计空间,通过可视化展示设计空间形态,能够提高固体火箭发动机优化设计效率和质量。采用试验设计方法对设计变量和约束条件重要性进行排序,简化优化模型;采用图形变形法对简化优化模型在不同试验设计点进行可视化分析;可视化选取可行设计空间中目标函数较小的点作为初始点。应用三种常规优化算法对4个不同初始点进行优化计算,计算结果表明,图形变形法优选的初始点使优化效率有不同程度的提高,优化结果具有更好的全局最优解特性;与遗传算法比较,效率提高400倍以上。
Graph Morphing is presented to explore the design space of solid rocket motor optimization problem. A formal procedure is presented to guide the designers to select a better initial point by Graph Morphing. First experiment design is used to simplify optimization model through ranking design variables and constraints. Then Graph Morphing is adopted to visualize the local region around the experiment design point of simplified model. Following that an initial point in feasible design space is selected which produces less objective value. Adopting three kinds of normal optimization algorithm to compute from four different initial point, it is demonstrated that beginning from the initial point selected by Graph Morphing can effectively reduce the computational time and improve the design quality. When comparing to computing with genetic algorithm, efficiency is improved over 400 times.
出处
《推进技术》
EI
CAS
CSCD
北大核心
2006年第3期201-204,207,共5页
Journal of Propulsion Technology
基金
国防预研基金项目资助(41328010504)
