期刊
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
卷 63, 期 3, 页码 903-926出版社
SPRINGER
DOI: 10.1007/s10589-015-9789-6
关键词
Black-box functions; Constrained global optimization; Kriging; Multi-objective optimization
Metamodeling, i.e., building surrogate models to expensive black-box functions, is an interesting way to reduce the computational burden for optimization purpose. Kriging is a popular metamodel based on Gaussian process theory, whose statistical properties have been exploited to build efficient global optimization algorithms. Single and multi-objective extensions have been proposed to deal with constrained optimization when the constraints are also evaluated numerically. This paper first compares these methods on a representative analytical benchmark. A new multi-objective approach is then proposed to also take into account the prediction accuracy of the constraints. A numerical evaluation is provided on the same analytical benchmark and a realistic aerospace case study.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据