Comparison of three space mapping techniques on electromagnetic design optimization. Issue 2 (5th March 2018)

• Record Type: Journal Article
• Authors: Gong, Jinlin; Gillon, Frédéric; Bracikowski, Nicolas
• Is Part Of: Compel. Volume 37:Issue 2(2018)
• Page Start: 565
• Page End: 580
• Publication Date: 2018-03-05
• Subjects: Kriging -- Optimal design -- 5-phase linear induction motor -- Manifold mapping -- Output space mapping; Electrical engineering -- Data Processing -- Periodicals; Electrical engineering -- Mathematics -- Periodicals; Electrical engineering -- Periodicals; Electronics -- Data Processing -- Periodicals; Electronics -- Mathematics -- Periodicals; 621.3
• DOI: 10.1108/COMPEL-11-2016-0512
• Languages: English
• ISSNs: 0332-1649

Abstract

Abstract : Purpose: This paper aims to investigate three low-evaluation-budget optimization techniques: output space mapping (OSM), manifold mapping (MM) and Kriging-OSM. Kriging-OSM is an original approach having high-order mapping. Design/methodology/approach: The electromagnetic device to be optimally sized is a five-phase linear induction motor, represented through two levels of modeling: coarse (Kriging model) and fine.The optimization comparison of the three techniques on the five-phase linear induction motor is discussed. Findings: The optimization results show that the OSM takes more time and iteration to converge the optimal solution compared to MM and Kriging-OSM. This is mainly because of the poor quality of the initial Kriging model. In the case of a high-quality coarse model, the OSM technique would show its domination over the other two techniques. In the case of poor quality of coarse model, MM and Kriging-OSM techniques are more efficient to converge to the accurate optimum. Originality/value: Kriging-OSM is an original approach having high-order mapping. An advantage of this new technique consists in its capability of providing a sufficiently accurate model for each objective and constraint function and makes the coarse model converge toward the fine model more effectively.