A multiobjective clustering of solutions for system reliability optimization

This article proposes a practical methodology to exactly solve the multiobjective redundancy allocation problem for series-parallel systems. A Pareto-optimal set is initially obtained by using a multiple objective evolutionary algorithm. The few studies that devoted to prune the size of the set utilized various approaches to classify the set. The clusters are compact, but they are not balanced. This article suggests a novel multiobjective clustering algorithm based on the notion of game theory, called game clustering which optimizes two conflicting objectives, named compaction and equi-partitioning. The definition of the payoff function considers both objectives with equal priority. A mixed strategy Nash equilibrium is performed by calculating probabilities corresponding to the strategies of the players. The game clustering approach can efficiently generate high performance and fairness clusters for all the Pareto-optimal solutions of multiobjective redundancy allocation problems. The experimental results are reported to demonstrate the efficiency of the proposed method.