Affiliations: [a] School of Mechanical and Electrical Engineering in Xuchang University, Xuchang, Henan, China
| [b] Faculty of Engineering in University of Ottawa, Ottawa, ON, Canada
| [c] National CAD Centre in Huazhong University of Science and Technology, Wuhan, China
Correspondence:
[*]
Corresponding author. Yizhong Wu, National CAD Centre in Huazhong University of Science and Technology, Wuhan, China. E-mail: [email protected].
Abstract: A Kriging-based global optimization method is proposed to solve black-box unconstrained design problems in this work. Firstly, the non-convex Kriging optimization problem is converted into the two convex programing problems by the canonical dual transform to quickly get global optimal solution. Then, PSO (Particle Swarm Optimization) algorithm is adopted to find next promising design point by exploring and optimizing the transformed problems. The proposed method not only reduces the computational burden, but also effectively balances local and global search behavior. Some well-known numerical test functions and a real engineering example are investigated to illustrate that the presented method can further enhance the feasibility, validity and robustness of the optimization process in contrast with other global optimization algorithms.
Keywords: Surrogate model, kriging, global optimization, dual transformation
