期刊
JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY
卷 29, 期 5, 页码 2121-2133出版社
KOREAN SOC MECHANICAL ENGINEERS
DOI: 10.1007/s12206-015-0434-1
关键词
Kriging-based algorithm; Expensive black-box functions; Global optimization; Local search strategy
资金
- National Natural Science Foundation of China [51375389]
- Fundamental Research Funds for the Central Universities [3102014JCQ01007]
In this paper, a novel kriging-based algorithm for global optimization of computationally expensive black-box functions is presented. This algorithm utilizes a multi-start approach to find all of the local optimal values of the surrogate model and performs searches within the neighboring area around these local optimal positions. Compared with traditional surrogate-based global optimization method, this algorithm provides another kind of balance between exploitation and exploration on kriging-based model. In addition, a new search strategy is proposed and coupled into this optimization process. The local search strategy employs a kind of improved Minimizing the predictor method, which dynamically adjusts search direction and radius until finds the optimal value. Furthermore, the global search strategy utilizes the advantage of kriging-based model in predicting unexplored regions to guarantee the reliability of the algorithm. Finally, experiments on 13 test functions with six algorithms are set up and the results show that the proposed algorithm is very promising.
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