An improved normal boundary intersection method for multiobjective inverse problems

Issue title: Selected Papers from the 15th International Symposium on Applied Electromagnetics and Mechanics (ISEM 2011)

Article type: Research Article

Authors: An, Siguang^{a; b} | Liu, Lei^a | Yang, Shiyou^{a; *}

Affiliations: [a] College of Electrical Engineering, Zhejiang University, Hangzhou, China | [b] Department of Electrical Engineering, China Jiliang University, Hangzhou, China

Correspondence: [*] Corresponding author: Shiyou Yang, College of Electrical Engineering, Zhejiang University, 310027, Hangzhou, China. Fax: +86 571 8795 1625; E-mail: [email protected]

Abstract: An improved normal boundary intersection (NBI) method is proposed. To overcome the deficiency of NBI methods in finding one Pareto solution in a single run, the utopia plane is divided into sub-domains. Dual external archives are proposed to enhance the diversity of the population. A new elite strategy is introduced to keep the balance between global and local searches. Two counter vectors are defined and used to record the mapping rate and selection rate of each sub-domain to act as elites to guide searches. The projection distance is employed to select candidates for next cycle of iterations. The expending of the ideal plane is implemented to sample completely the whole objective space and obtain an even distributed Pareto front. To validate the proposed algorithm, it is used to solve mathematical functions and an inverse problem with promising results.

Keywords: Normal boundary intersection method, inverse problem, multi-objective, evolutionary algorithm

DOI: 10.3233/JAE-2012-1451

Journal: International Journal of Applied Electromagnetics and Mechanics, vol. 39, no. 1-4, pp. 121-127, 2012