Reliability Logistics Network Design Based on Two Stage Robust Optimization

This paper investigates the reliability logistics network design problem with uncertainty due to node and road damages. A reliability logistics network design method based on two-stage robust optimization is proposed. Taking the location of supply and transit nodes, determination of node connectivity, and distribution of cargo flow as decision variables, a two-stage reliability logistics network design model was constructed. The model has two independent objectives, i.e., the total network cost objective function and the total network operation time objective function. The former includes two stages of cost, in which the first stage calculates the network construction cost and the network operation cost under the normal state and the second stage calculates the network operation cost under a disruptive scenario set. The total network operation time objective function is used to calculate the network operation time under the normal state. A hybrid evolutionary algorithm with double-layer encoding structure chromosomes is designed. The NPGA is used as the main algorithm framework, and a large neighborhood search mechanism is designed to optimize the connectivity relationship genes of individuals. Cluster-based crossover and mutation strategies are combined to improve the search ability of the algorithm in the solution space. The effectiveness of the model and algorithm are verified by case studies with several groups of reliability logistics network design problems of different scales. The results show that the two-stage reliability logistics network design model can significantly reduce the operation cost of the network in case of damage and effectively improve network reliability through a small increase in the initial network construction cost. In the case comparison of 5 supply nodes, 10 transit nodes, and 15 demand nodes, the two sets of Pareto solutions for cost preference and time preference obtained by the model can save up to 20.6% and 28.2% of network operation cost in the same network damage scenario set, respectively, compared with the two sets of corresponding preference solutions of the traditional multi-objective logistics network model. The hybrid evolutionary algorithm converges to a better target value at the initial stage of iteration, showing a better search and optimization performance, which can effectively solve the two-stage reliability logistics network design model.