Redundancy allocation and optimal buffer size determination in series-parallel production systems: A hybrid approach

Buffer allocation and redundancy allocation problems arose during the design of the production line. This study investigated these two problems as an integer nonlinear multi-objective model. The purposes of the model included maximizing availability and minimizing overall system costs and buffer capacity. Machine failures were divided into preventive and emergency categories, and their costs were factored into the cost goal function. Furthermore, each machine's failures were considered accidental and followed Weibull distribution functions. To solve the problem, a method combining the design of experiments, simulation, neural network, and meta-heuristic algorithm was applied. The system was simulated using the results of designed experiments, and the results in neural networks were analyzed to estimate the functions related to cost and availability purposes. Finally, the neural network's estimated functions, another objective function (buffer capacity), and the problem's constraints were implemented in the meta-heuristic non-dominated sorting genetic algorithm III (NSGA-III) to obtain Pareto sets. The results were compared to the non-dominated sorting genetic algorithm II (NSGA-II) and the Multi-objective particle swarm optimization algorithm (MOPSO). The applicability of the proposed approach was demonstrated using a real case, and the outcomes validated the algorithm's effectiveness.