An integrated approach to solve a robust forward/reverse supply chain for short lifetime products

• An integrated robust supply chain problem for short lifetime products is developed.
• A cyclic feedback procedure is proposed in between location and allocation models.
• A new method is developed to linearize the mixed integer nonlinear programming model.
• DE outperforms TS algorithm at the expense of longer running time.

Supply chain management plays a key role in optimizing manufacturing costs in competitive markets. In this paper, a multi period, multi product, multi echelon and capacitated supply chain problem for short lifetime products was solved using both exact method and metaheuristic algorithms. The considered echelons in a forward/reverse supply chain network design include suppliers, hybrid production–inspection centers, hybrid warehouse-collection centers, retailers, disposal centers and recovery centers. To tackle both strategic and tactical nature of the problem and due to its complexity, the location problem is first solved using an exact method in which strategic decisions are made aiming to minimize the location costs. Then the allocation problem is solved using both exact and metaheuristic algorithms, namely Tabu Search and Differential Evolution, in which based on the results obtained from the location problem, the profit is maximized and tactical decisions are made. A cyclic feedback procedure is proposed in between the location and allocation models to obtain the best solution. The allocation model is linearized using a new method and to handle the uncertainty, a robust mixed integer programming model is developed which can retain the linearity of the model. The computational results confirm the effectiveness of the exact method for the location problem while testifies the proposed algorithms can obtain effective solutions compared with that of the exact solutions for the allocation problem in both deterministic and uncertain environment. In addition, the results show that in terms of preciseness and convergence, the proposed Differential Evolution algorithm outperforms the other algorithm.