多重元模型搜索方法及其在汽车轻量化设计中的应用

目前广泛应用的基于元模型的优化方法通常起始于单一一组样本点,一旦样本点质量较差、其搜索效率和精度都会受到影响,甚至得不到想要的结果。提出一种多重元模型搜索方法,此方法应用两组初始样本点,拟合多个元模型。在搜索过程中,首先应用部分昂贵点构造一个重点空间,并构建二阶多项式进行搜索。为避免陷入局部最优的陷阱,同时应用构建的克里金模型搜索整个设计空间。经过多个算例的测试,此方法具有很高的搜索精度、搜索效率和稳健性。将此方法应用于某款车的轻量化设计中,经过优化,此结构减少了8.0 kg的质量,并提升了其刚度性能求。

Multiple Sets of Meta-Model Based Search Method and Its Applications to Vehicle Lightweight Design

The meta-model based methods are often used to solve time-consuming problems in design optimization. The existing meta-model based algorithms usually include one set of initial points. However,the low quality initial points may lead to the decreased performance and undesired results. In this work,a search method based on multiple sets of meta-model was proposed. The method starts with two different sets of initial points to fit multiple meta-models. In the search process,some of the expensive points were used to construct an important region,which would then be searched through using the fitted QF models. The Kriging model was used in the search process to avoid being trapped into local optima. Tested by several math functions,the proposed method showed excellent search accuracy,efficiency and robustness. The method was then applied in a vehicle lightweight design. The results show that 8.0 kg of vehicle weight is reduced when the stiffness increases.