A local supply chain inventory planning with varying perishability rate product: A case study

Abstract

Motivated by a real-life case in the dairy supply chain, we study the management of deteriorating inventories in an existing three-echelon supply chain network with a local retailer with multiple customers. Like many industries, the focal company categorizes the customers into two different clusters in the supply chain downstream with different, uncertain demand functions: (i) many, small, ordinary customers; (ii) few, large, premier customers. The focal company intends to determine the optimal inventory policy while the deterioration rate of the product is time sensitive. The problem is formulated using two concepts: 1) queuing theory and finite-horizon Semi-Markov process to characterize the structure of the optimal inventory policy; 2) integrated inventory system in the supply chain at the network level. We formulate the problem mathematically and show that the model is nonlinear and nonconvex. Therefore, the model cannot be solved by using commercial optimization software products to reach global optimal. Consequently, two solution approaches are designed to solve the problem: (i) a modified version of the Invasive Weed Optimization algorithm, named adaptive Invasive Weed Optimization, which is designed, tuned and validated (using well-known algorithms) to tackle the problem; (ii) an Adaptive Heuristic Method. We show that adaptive Invasive Weed Optimization is superior in the quality of solutions while Adaptive Heuristic Method is faster. Various sensitivity analyses applied to the real data show when facing both perishability and uncertainty, the effect of perishability outweighs the effect of uncertainty.