摘要
通过引入非优超排序和排挤的多目标处理机制 ,将分布式协同进化MDO算法的能力扩展到多目标的多学科设计优化问题。多目标的分布式协同进化MDO算法在保持各学科充分自治和各学科并行设计优化协同的基础上 ,通过一次运行即可获得具有良好分布的多个Pareto最优解 ,逼近整个Pareto最优前沿。应用于导弹气动 /发动机 /控制三学科两目标设计优化问题 ,与约束法计算结果的对比表明算法能够有效逼近该问题的Pareto最优前沿 。
By introducing multiobjective handling mechanism of nondominated sorting and crowding, ability of distributed coevolutionary multidisciplinary design optimization algorithm is extended to multiobjective multidisciplinary design optimization (MDO) problems. The multiobjective distributed coevolutionary MDO approach maintains sufficient disciplinary autonomy, exploits synergism of disciplinary concurrent design optimizations, meanwhile it obtains a set of well distributed Pareto optimal solutions and makes a good approximation of the whole Pareto optimal front in one single run. It is applied to a missile design optimization problem with two system objectives and three disciplines-aerodynamic, engine, and control. Comparison with the results of the constraint method indicates that the multiobjective coevolutionary MDO approach can effectively approximate Pareto optimal front of the ploblem, providing plenty of information for design decision making.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2002年第4期12-15,共4页
Journal of National University of Defense Technology
关键词
MDO算法
导弹设计
多学科设计优化
多目标优化
进化计算
协同进化算法
分布式计算
missile design
multidisciplinary design optimization
multiobjective optimization
evolutionary computation
coevolutionary algorithms
distributed computing
