摘要
作为一种贝叶斯优化算法,高效全局优化算法(EGO)利用克里格模型来构造近似模型,并采用样本填充准则以寻找下一个样本点来更新近似模型。文中详细介绍了该优化算法,并将其应用于船舶力学的两个典型优化例子。其中一个是潜艇的多学科概念设计,考虑了水动力、推进、重量、性能和成本5个学科;另外一个是屈曲状态下加筋板的优化问题。与传统优化相比,高效全局优化算法不仅收敛到全局最优解,而且更加有效。结果表明高效优化算法非常适用于船舶力学中的优化问题。
Efficient Global Optimization (EGO), one Bayesian analysis optimization algorithm, makes use of Kriging model to construct statistical approximation model and uses infill sampling criteria (ISC) to find the next sampling point updating the model. This method is discussed in detail and applied in ship mechanics with two traditional optimization examples.One is a submarine conceptual multidisciplinary design, which considers hydrodynamics, propulsion,weight and volume, performance, and cost. It is a mixed-variable optimization problem defined by 8 real design variables,3 integer design variables and 12 constraints.The other is stiffened panel optimization under buckling.Compared with traditional methods, EGO not only finds the global optimal point, but also completes the optimization more efficiently. The results demonstrate that EGO is very suitable for optimization in ship mechanics.
出处
《船舶力学》
EI
北大核心
2008年第3期473-482,共10页
Journal of Ship Mechanics