利用模型管理框架求解多目标优化问题

Solving multi-objective optimization problem using model management framework

为了提高演化算法的效率,减少优化时间,提出一种多目标模型管理框架。利用该模型管理框架可以在整个寻优区域内建立比较精确的目标及约束的近似模型,从而避免了大量耗时的高精度分析计算。将该多目标模型管理框架与单纯形-多目标粒子群算法(SM-MOPSO)相结合,对某轻型飞机齿轮箱减速器进行多目标优化设计,使高精确分析计算的次数减少88%。该多目标模型管理框架及SM-MOPSO算法可用于求解大型、复杂的工程优化问题。

In order to improve the efficiency of evolutionary algorithm and reduce the optimization time, a multi-objective model management framework was proposed. By using the multi-objective model management framework, accurate approximation models of the entire searching space can be constructed, and evolutionary algorithm will avoid a great number of time-consuming highfidelity analyses. The multi-objective model management framework is integrated with a new hybrid evolutionary algorithm, simplex method-multiple objective particle swarm optimization （SM-MOSPO）, to solve multi-objective optimization design of speed reducer gearbox. Not only a good Pareto set is obtained efficiently, but also the time of high-fidelity analyses is reduced by 88%. In complex engineering optimization designs, the multi-objective model management framework and SM-MOSPO algorithm should be recommended.