Subdivision-based optimal design of multi-dimensional electromagnetic devices

Abstract

Until now, in the field of electrical engineering, there have been many contributions on optimal design of electromagnetic devices. The gradient based methods such as the steepest descent method have fast convergence ability, however, they are apt to be trapped into a local optima [1]. Compared with gradient based methods, the evolutionary optimization algorithm have strong global searching ability, but it needs more particles and iterations [2]. For large scale electromagnetic problem, the expensive performance analysis is also a big challenge for optimization. Even the surrogate model, for example, the Kriging technique, has been introduced to assist the optimization process, to find the true global optimal design, it still needs a large number of sampling designs [3]. Based on the above background, it is urgent to develop one effective optimal design algorithm, which can guarantee a fast searching for the global optimal design especially for complex engineering problems. In the field of structural analysis, the substructure methods have been used extensively in the analysis of complex structures by splitting them into substructures [4]. The same method, with all its computational savings and merits, may equally well be employed in electromagnetic field analysis.