Inventory Optimization With Resource Constraint: An Application

The following document presents the development and results of a researchproject conducted by the Department of Industrial Engineering of Tsinghua Universityand sponsored by Company, Ltd. It started in July2010and aimed at optimizing theinventory and operation cost of one of Company, Ltd warehouses. The companyprovided historical demand data, detailed description of the warehouse as well asprecise characteristics of all parts stocked in the warehouse. The main requirements ofthe project were to determine the precise parameters of each part’s inventory controlpolicy and the exact placement of each stock in the warehouse. This document focusessolely on the inventory cost optimization aspect of the project.Optimizing inventory cost consists more specifically in optimizing the holdingcost and set-up cost of each part in the warehouse. In our project (r,Q) policies are usedto control and optimize inventories. The first step of our solution approach was tothoroughly analyze the historical demand data. Χ2tests were conducted to evaluatewhether certain parts’ demand distributions followed a compound Poisson processdistribution. The parts were then sorted into three different types. Type1parts haveonly one demand point in their historical data. Type2parts have several demandpoints in their historical demand records and a demand distribution which does notfollow a compound Poisson distribution. Type3parts also have several demand points,however unlike type2, their demand distribution follows a compound Poissondistribution.One important specificity of our optimization problem is the presence of storage-space constraints. On top of this storage-space constraint, a type I service levelobjective has to be respected. For each type of parts, particular solution methods weredeveloped. Because of the lack of historical data, the solution approach for parts withone demand point is deterministic. Type2parts approach is based on the discretedistribution of the demand during a lead-time as well as on the Economic OrderQuantity model while type3parts approach uses the closed form of the inventory costwhen the demand follows a compound Poisson process. Finally, our simulation results are analyzed. The algorithm offers convincingresults of cost-reduction ranging from5to40%depending on the service level,holding cost and set-up cost coefficients.