Multi-start simulated annealing approach for warehouse network redesigning problem with zone dependent fixed cost

Abstract

The warehouse location model is one of the important researches in the location analysis field, but it is different from other types of location problems such as the facility location problem and also the location and allocation problem since it is strongly affected by the current economic situation. In addition, existing warehouse network redesign models and mathematical programming models do not consider the consolidation, elimination, and addition of a new warehouse simultaneously in one model but only consider one or a combination of two of them only. However, these three redesign techniques namely addition, closure, and consolidation of a new warehouse should be considered in the model to optimize the cost. Therefore, this study aims to construct a warehouse network model that minimizes the total supply chain costs by consolidating, eliminating, and adding new sites. It is done through the implementation of Simulated Annealing (SA) optimization method that adopts cooling schedule variants, dynamic stopping criteria, and multi-start. A study of SA parameters such as cooling strategies, initial temperature, and stopping criteria is conducted to ensure a more efficient SA procedure. The study is carried out on three types of network data types: data type 1 – uniform, data type 2 – cluster, and data type 3 – compact. The performance comparisons are done among three different SA cooling schedules: geometric, logarithmic, and linear on the three different sets of data. Dynamic stopping criteria has been chosen such that the algorithm will stop after finding five consecutive best solutions or when there was no improvement in the cost after hundred successive iterations. The computational experiments showed that the geometric cooling schedule produced consistently better-quality solutions in a shorter time than the other schemes’ solution. The best initial temperature was found to vary for all three sets of data. This research found that when the number of multi starts increases, the average cost is also proportionally decreasing. On the additional redesign network factors such as the zone dependent fixed cost and capacity constraint, the results showed that when the cost of the zone is introduced, the model suggested the warehouse to be opened in low-value zones. Warehouses in the high-value zone will be closed, and its operations combined with nearby warehouses in the low-value zone. Consequently, warehouses that are in the low-value zone will usually receive a lot of customers from the high-value zone causing increased merger costs for nearby warehouses. The capacity constraint is also considered in the model to control the number of customers that need to be supplied by a warehouse so that the model will represent the real problem more closely. With the addition of this constraint, the customers are not necessarily served by the nearest warehouse hence will increase the transportation cost.