一种基于近似模型管理的多目标优化方法

针对实际工程优化问题,为了提高效率和减少近似模型带来的误差,提出一种基于模型管理的多目标优化方法。利用加强径向基插值函数在整个寻优区域内构造目标和约束的近似模型,结合微型多目标遗传算法寻找当前非支配解。通过模型管理方法更新近似模型,并控制由于近似模型带来的误差和更新次数,最后将误差控制在一定范围内的多个非支配解当作实际问题的解。在测试函数中验证了此方法的效率及非支配解的精度和分布的均匀性。最后成功应用于车身薄壁构件的耐撞性优化中,表明了可用于求解复杂的工程优化问题。

A MULTI-OBJECTIVE OPTIMIZATION METHOD BASED ON THE APPROXIMATE MODEL MANAEMENT

A multi-objective optimization method based on model management framework is suggested to improve the efficiency and decrease the error of the approximate model for the computation-intensive engineering problem. The approximate model is built by extended radial basis functions,and then the non-dominated solutions predicted by the approximate model are identified through the micro multi-objective genetic algorithm. Then,the management framework updates the approximate model and controls the iteration time and the error resulting from the approximate model. Finally,the non-dominated solutions whose errors are controlled appropriately can be regarded as solutions to the problem. The proposed method is successfully applied to the analysis of thin-walled sections for structural crashworthiness,demonstrating that it can solve complex computation-intensive engineering multi-objective optimization problem.