Constrained design optimization of active magnetic bearings through an artificial immune system. Issue 8 (7th November 2016)

• Record Type: Journal Article
• Authors: Chou, Yu-Cheng; Fan, Yi-Hua; Nakajima, Madoka; Liao, Yi-Lin
• Is Part Of: Engineering computations. Volume 33:Issue 8(2016)
• Page Start: 2395
• Page End: 2420
• Publication Date: 2016-11-07
• Subjects: Artificial immune system -- Design optimization -- Constrained optimization -- Active magnetic bearing; Computer-aided engineering -- Periodicals; Computer graphics -- Periodicals; 620.00285
• DOI: 10.1108/EC-05-2015-0137
• Languages: English
• ISSNs: 0264-4401

Abstract

Abstract : Purpose: The purpose of this paper is to present the use of artificial immune systems (AISs) to solve constrained design optimization problems for active magnetic bearings (AMBs). Design/methodology/approach: This research applies the AIS approach, more specifically, a representative clonal selection-based AIS called CLONALG, to the single-objective structural design optimization of AMBs. In addition, when compared with a genetic algorithm (GA) developed in the previous work, the CLONALG fails to produce best solutions when a nearly zero feasible ratio occurs in an AMB design problem. Therefore, an AIS called ARISCO (AIS for constrained optimization) is proposed to address the above issue. Findings: A total of six AMB design cases are solved by the GA, CLONALG, and ARISCO. Based on the simulation results, in terms of solution quality, the ARISCO is shown to have better overall performance than the CLONALG and GA. In particular, when solving a problem with a nearly zero feasible ratio, the ARISCO and GA perform equally and both outperform the CLONALG. Originality/value: In summary, the contributions of this paper include: this research applies the AIS approach, more precisely, the CLONALG, to the single-objective structural design optimization of AMBs; the ARISCO overall produces better AMB designs than the CLONALG and a GA developed in the previous work; in situations where a nearly zero feasible ratio occurs, the ARISCO and GA perform equally, and they both outperform the CLONALG.