Genetic search for redundancy optimization in complex systems

Abstract

PurposeConceiving reliable systems is a strategic issue for any industrial society for its economical and technical development. This paper aims to focus on solving highly constrained redundancy optimization problems in complex systems.Design/methodology/approachGenetic algorithms (GAs), one of the metaheuristic techniques, have been used and a dynamic adaptive penalty strategy is proposed, which makes use of feedback obtained during the search along with a dynamic distance metric and helps the algorithm to search efficiently for final, optimal or near optimal solution.FindingsThe effectiveness of the adaptive penalty function is studied and shown graphically on the solution quality as well as the speed of evolution convergence for several highly constrained problems. The investigations show that this approach can be powerful and robust for problems with large search space, even of size 1017, and difficult‐to‐satisfy constraints.Practical implicationsThe results obtained in this paper would be applicable on designing highly reliable systems meeting the requirement of today's society. Moreover, an important advantage of applying GA is that it generates several good solutions (mostly optimal or near optimal) providing a lot of flexibility to decision makers. As such, the paper would be of interest and importance to the system designers, reliability practitioners, as well as to the researchers in academia, business and industry. The paper would have wide applications in the fields of electronics design, telecommunications, computer systems, power systems etc.Originality/valueGenetic algorithms have been recently used in combinatorial optimization approaches to reliable design, mainly for series‐parallel systems. This paper presents a GA for parallel redundancy optimization problem in complex systems.