基于代理模型和等效刚度模型的加筋柱壳混合优化设计

针对轴压作用下的加筋柱壳后屈曲性能优化计算较大的问题,本文提出了一种基于代理模型和等效刚度模型的混合优化策略,即运用基于等效刚度的平铺模型进行有限元后屈曲分析以代替试验设计中大量的精细加筋模型分析,并通过控制等效模型的单元尺寸来调整其分析精度,而等效刚度模型计算时长仅约为精细加筋模型的1/3。对构建的代理模型采用多岛遗传算法进行极限承载力等约束下的轻量化设计,调用精细模型有限元后屈曲分析对代理模型进行更新,从而保证代理模型的拟合精度并得到优化解。工程算例结果表明,本文提出的混合优化方法,使加筋柱壳结构在满足承载力情况下减重效果明显。

Hybrid optimization of grid-stiffened cylinder based on surrogate model and smeared stiffener model

With regard to the grid-stiffened cylinder under non-uniform axial compression,it is time-consuming to perform the post-buckling optimization.A hybrid optimization approach based on surrogate model and smeared stiffener model is proposed to solve this problem.First,smeared stiffener model is used to carry out the design of experiment instead of exact grid-stiffened model,in which a high fitting precision can be reached by adjusting the element size.Meanwhile,the computational time is reduced to one-third of the exact model.Then,surrogate model is built for optimization using multi-island genetic algorithm to obtain the minimum mass solution under buckling constraints.After several iterations,the analysis of exact model is utilized to update the surrogate model.Finally,an engineering example is established to illustrate the proposed approach,and the results indicate that the hybrid optimization can gain a significant weight reduction with less computational effort.