C Conferentia Proceedings
CILAMCE2015-0631 COMPUTATIONAL INTELLIGENCE TECHNIQUES FOR OPTIMIZATION AND DATA MODELING

Handling optimization problems with constraints of different magnitudes using evolutionary algorithms

Rafael de P. Garcia1; Beatriz S. L. P. Lima1; Breno P. Jacob1; Afonso C. de C. Lemonge2

1 Federal University of Rio de Janeiro; 2 Federal University of Juiz de Fora

doi:10.20906/CPS/CILAMCE2015-0631

Resumo

This paper proposes a rank based constraint handling technique called Multiple Constrained Ranking Technique (MCR). This technique is a modification of the ranking method by Ho-Shimizu (HS), where a population of individuals is sorted using three ranks as following: the value of the objective function, the squared sum of the violations values and the number of violated constraints. Our proposal changes the second rank, which is the sum of the violations values by considering many ranks separately, one for each constraint of the problem. The aim of this new constraint handling technique is to overcome the difficulty shown by other techniques when faced with constraint violation values of different magnitudes. We employ the MCR technique with a genetic algorithm using the 2006 CEC benchmark functions and some well-known engineering problems. The results suggest that the MCR technique is competitive, especially when applied to problems defined by constraints of very different magnitudes.

Palavras-chave: Constrained ranking technique; Evolutionary algorithms

Como citar

Rafael de P. Garcia; Beatriz S. L. P. Lima; Breno P. Jacob; Afonso C. de C. Lemonge. “Handling optimization problems with constraints of different magnitudes using evolutionary algorithms”. XXXVI Ibero-Latin American Congress on Computational Methods in Engineering. CILAMCE2015. 2015. DOI: 10.20906/CPS/CILAMCE2015-0631