Abstract:

Robust optimization is time-consuming for uncertainty analysis. Aiming to improve computing efficiency, a sequential robust optimization method was proposed, which combined dual-Kriging model with multi-point sampling criterion. Dual-Kriging model, constructing the relationship between variables and standard deviation of the objective, reduced the computing time of uncertainty analysis in optimization to improve computing efficiency. Multi-point sampling criterion realized sequential iteration process, and improved the global optimum. The test results of a mathematical function showed that the proposed robust optimization method was more efficient in comparison with common surrogate-based robust optimization. Lastly, the proposed optimization method was applied to optimizing the wall-thickness and injection process for a box-shape injection molded part by considering the instability of injection process, and the results showed that the optimization method was effective.

摘要：

稳健优化设计需要进行不确定性分析，优化过程比较费时。为提高计算效率，提出了基于Kriging代理模型的稳健优化设计方法。该方法结合双重Kriging代理模型和多点加点准则进行稳健优化设计。双重Kriging代理模型的建立，减小了不确定性分析的计算量，提高了计算效率。多点加点准则实现了序列的迭代优化，在保证计算效率的同时提高了对全局最优解的逼近程度。数学函数测试结果表明，与一般模拟基稳健优化方法相比，本文所提出的序列稳健优化方法在效率上有很大的提高。最后，考虑注塑工艺的不稳定性，对一盒式注塑制件的壁厚和工艺进行稳健优化，结果表明该稳健优化设计方法是有效的。