Enhanced butterfly optimization algorithm for reliability optimization problems

In this study a recently introduced algorithm, based on foraging process of butterflies known as butterfly optimization algorithm (BOA) is explored and a modified variant is introduced. The framework of BOA is based on the fragrance emitted by the butterflies, which helps other butterflies in searching food as well as in identifying a mating partner. BOA performs both the local and global search while seeking for the global optimal solution for the problem. Despite this BOA sometime stuck in a local optima which results in a slow or poor convergence. This study embeds the bidirectional search in the structure of BOA. This helps to perform the local search in forward as well as backward direction. Greedy selection is performed, while selecting the direction. If the solution improves while traversing backward then backward traverse is adapted otherwise forward. The proposed variant is termed as bidirectional butterfly optimization algorithm (BBOA). This modification facilitate in accelerating the convergence rate of BOA, which is verified and validated by statistical and comparative results on a set of CEC2006 and CEC2014 benchmark problems. Non-parametric statistical tests are performed to analyze the results. Further the method is investigated to solve eight reliability optimization problems. Experimental results demonstrate the competitiveness of BBOA.