Development of integrated models for distribution network design of perishable products

Over the last few years, inventory-location network design models due to the incorporation of two main decisions in network design, meaning inventory control and facility location, has attracted great attention from both industries and researchers. Existing inventory-location models are mainly applicable for nonperishable inventories. However, since, perishable inventory comprise a large proportion of the inventory distributed through the distribution networks, and due to the important role of the network design on the final cost and quality of the products, development of inventory-location models suitable for perishable inventory is of great importance. Therefore, the objective of this research is development of integrated inventory-location models for distribution network design of fixed- and random-lifetime perishable products, that has been achieved through the following four phases. The first phase is to select the best structure for the distribution network between a centralized and decentralized configuration. Previous research works that admire a centralized structure consisting of suppliers, distribution centers and retailers, considered for simplicity that the products dispatched from the supplier are maintained at the distribution centers locations before being demanded by the customer. It is equivalent to this assumption that if a customer places an order to a retailer, and the retailer is out-of-stock the customer waits for the products to be available. However, in practice it is not true and a fraction of customers encountering stock-out switches to other markets. To this end, the first part of this study formulates the centralized and decentralized structures when inventory can be stored at both retailers’ and DCs’ location. Sensitivity analysis is conducted to determine the value of centralization. Results indicate that the value of centralization is dependent on the parameters of the problem, and that for the parameters of the problem considered in this research, a centralized structure has been more efficient. So, for the rest of this research a centralized structure is considered for the distribution network. The second phase of this research is to determine the best optimization approach to formulate the problem between an integrated and a decoupled method. Hence, the inventory-location model is formulated for fixed-lifetime perishable product once by an integrated model and another time by a decoupled model. A memetic algorithm (MA) is developed to solve the integrated model, and the Lingo software is applied to solve the decoupled model. Results show that up to 5.7% cost reduction is achieved by applying the integrated model. The developed MA is compared with a GA from literature. Result indicates that in terms of solution quality, MA is up to 22.13% superior to GA, but up to 4.17 times slower than that. Therefore, the third phase of this research develops a much efficient solution method based on Lagrangian relaxation to solve the integrated inventory-location model for fixedlifetime perishable products. Results show that the Lagrangian relaxation algorithm is in average 255 times faster than GA, and in terms of solution quality is up to 22.13% better than GA. The developed model, other than facility location and inventory control decision for a distribution network, provides the managers the opportunity to select among higher-inventory-cost options that lead to longer-lifetime for products and less costly options that result in having products with shorter lifetimes. Finally, the last phase of this study develops an integrated inventory-location model for the network design of random-lifetime perishable products. The model defines the randomness of product lifetimes by discrete scenarios, and therefore, provides solutions that perform well under all defined scenarios. A Lagrangian relaxationbased heuristic algorithm is developed to solve the model. The algorithm produces solutions that are within 0.027% of optimality gap.

Preview Text FK 2015 146IR.pdf Download (938kB) | Preview

Actions (login required)

View Item