Aerodynamic optimization using a parallel asynchronous evolutionary algorithm controlled by strongly interacting demes

Asouti, VG; Giannakoglou, KC

dc.contributor.author	Asouti, VG	en
dc.contributor.author	Giannakoglou, KC	en
dc.date.accessioned	2014-03-01T01:29:48Z
dc.date.available	2014-03-01T01:29:48Z
dc.date.issued	2009	en
dc.identifier.issn	0305-215X	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/19350
dc.subject	asynchronous evolutionary algorithm	en
dc.subject	parallelization	en
dc.subject	aerodynamic shape optimization	en
dc.subject	multi-objective optimization	en
dc.subject.classification	Engineering, Multidisciplinary	en
dc.subject.classification	Operations Research & Management Science	en
dc.subject.other	GENETIC ALGORITHMS	en
dc.subject.other	DIFFERENTIAL EVOLUTION	en
dc.subject.other	DESIGN	en
dc.title	Aerodynamic optimization using a parallel asynchronous evolutionary algorithm controlled by strongly interacting demes	en
heal.type	journalArticle	en
heal.identifier.primary	10.1080/03052150802415665	en
heal.identifier.secondary	http://dx.doi.org/10.1080/03052150802415665	en
heal.language	English	en
heal.publicationDate	2009	en
heal.abstract	A parallel asynchronous evolutionary algorithm controlled by strongly interacting demes for single- and multi-objective optimization problems is proposed. It is suitable even for non-homogeneous, multiprocessor systems, ensuring maximum exploitation of the available processors. The search algorithm utilizes a structured topology of evaluation agents organized in a number of inter-communicating demes arranged on a 2D supporting mesh. Once an evaluation terminates and a processor becomes idle, a series of intra- and inter-deme processes determines the next agent to undergo evaluation on this specific processor. Real coding and differential evolution operators are used. Mathematical and aerodynamic-turbomachinery optimization problems are presented to assess the proposed method in terms of CPU cost, parallel efficiency and quality of solutions obtained within a predefined number of evaluations. Comparisons with conventional evolutionary algorithms, parallelized based on the master-slave model on the same computational platform, are presented.	en
heal.publisher	TAYLOR & FRANCIS LTD	en
heal.journalName	ENGINEERING OPTIMIZATION	en
dc.identifier.doi	10.1080/03052150802415665	en
dc.identifier.isi	ISI:000263644100003	en
dc.identifier.volume	41	en
dc.identifier.issue	3	en
dc.identifier.spage	241	en
dc.identifier.epage	257	en