A hybrid meta-heuristic approach for buffer allocation in remanufacturing environment

Abstract

Remanufacturing system is complicated due to its stochastic nature. Random customer demand, return product rate and system unreliability contribute to this complexity. Remanufacturing systems with unreliable machines usually contain intermediate buffers which are used to decouple the machines, thereby, reducing mutual interference due to machine breakdowns. Intermediate buffers should be optimized to eliminate waste of resources and avoid loss of throughput. The Buffer Allocation Problem (BAP) deals with allocating optimally fixed amount of available buffers to workstations located in manufacturing or remanufacturing systems to achieve specific objectives. Optimal buffer allocation in manufacturing and remanufacturing systems not only minimizes holding cost and stock space, but also makes facilities planning and remanufacturing decisions to be effectively coordinated. BAP in a non-deterministic environment is certainly one of the most difficult optimization problems. Therefore, a mathematical framework is provided to model the dependence of throughput on buffer capacities. Obviously, based on the survey undertaken, not only there exists no algebraic relation between the objective function and buffer size but the current literature does not offer analytical results for buffer capacity design in remanufacturing environment. Decomposition principle, expansion method for evaluating system performance and an efficient hybrid Meta-heuristic search algorithm are implemented to find an optimal buffer allocation for remanufacturing system. The proposed hybrid Simulated Annealing (SA) with Genetic Algorithm (GA) is compared to pure SA and GA. The computational experiments show better quality, more accurate, efficient and reliable solutions obtained by the proposed hybrid algorithm. The improvement obtained is more than 4.18 %. Finally, the proposed method is applied on toner cartridge remanufacturing company as a case study, and the numerical results from hybrid algorithm are presented and compared with results from SA and GA.