Decomposition-based evolutionary algorithm for large scale constrained problems Eman Sayed ⇑ , Daryl Essam, Ruhul Sarker, Saber Elsayed School of Engineering and Information Technology, UNSW-Canberra, Canberra, Australia article info Article history: Received 20 November 2013 Received in revised form 8 September 2014 Accepted 12 October 2014 Available online 24 October 2014 Keywords: Large scale constrained problem optimization Decomposition approach Variable interaction identification abstract Cooperative Coevolutionary algorithms (CC) have been successful in solving large scale optimization problems. The performance of CC can be improved by decreasing the number of interdependent variables among decomposed subproblems. This is achieved by first identifying dependent variables, and by then grouping them in common subproblems. This approach has potential because so far no grouping technique has been mainly developed for constrained problems. In this paper, a new variable interaction identification technique to identify the dependent variables in large scale constrained problems is proposed. The proposed technique is tested on both a new test suite of constrained problems with med- ium and high dimensions, which include overlapping subproblems and different levels of complexity and nonseparability and also the established DED problem. The experimental results have shown that the proposed technique contributes to the decomposition approach over a range of high dimensions, in comparison with other state-of-the art group- ing techniques. It achieves better performance with higher feasibility ratios and less com- putational time. Ó 2014 Elsevier Inc. All rights reserved. 1. Introduction Many real life problems (i.e. industry [44,69], environment [45], bio-computing [4], shape design (such as turbine blades [78], aircraft wings [87], and heat exchangers [26]), and operations research (such as optimization of water management problem [35], optimization of Norwegian natural gas production and transport [71], optimizing US army stationing [18]) are large scale optimization problems. The increased need for high quality decision making for such problems, has made the topic of solving large scale optimization problems a valuable and challenging research area. Moreover, the challenge increases when solving large scale constrained optimization problems. Although evolutionary algorithms (EAs) are an effec- tive optimization technique [40], the performance of EAs eventually deteriorates with increasing dimensionality [64]. Apply- ing a decomposition approach can reduce the dimensionality problem of EA by decomposing a large scale problem into smaller subproblems. However, to be more effective, before starting the optimizing process an EA should use a technique to identify and group the dependent variables of the large problem into smaller scale subproblems, in a way that decreases the interdependency among the subproblems. Using a decomposition approach for large scale unconstrained problems is not new in the literature. The logic of under- standing the complexity of systems through decomposition is found in the publications of Descartes and Veitch [19] and http://dx.doi.org/10.1016/j.ins.2014.10.035 0020-0255/Ó 2014 Elsevier Inc. All rights reserved. ⇑ Corresponding author. E-mail addresses: e.hasan@student.adfa.edu.au (E. Sayed), d.essam@adfa.edu.au (D. Essam), r.sarker@adfa.edu.au (R. Sarker), s.elsayed@adfa.edu.au (S. Elsayed). Information Sciences 316 (2015) 457–486 Contents lists available at ScienceDirect Information Sciences journal homepage: www.elsevier.com/locate/ins