Dynamic Scheduling Policies in Production-Inventory Systems with Returns or Two Classes of Demands

Abstract

A major concern in operations and supply chain management research has been to determine optimal policies to minimize costs in production and inventory systems. This dissertation consists of two applications of production-inventory systems motivated by the real world observations. The first model is motivated by a high-tech company data center server inventory control process. The firm has internalized the production process and faces server returns at the same time. There are demands for both new and used products. Returned servers can be used to serve the demand for use products directly, or be remanufactured into new condition by the same production capacity. Unmet demands are lost. The objective is to determine the optimal scheduling policy that minimizes the discounted total cost. We model the system with a production-inventory model with returns and show the optimal production inventory policy is of base stock type. Efficient heuristics are proposed and the base model is extended to scenarios that better described the real world situation. These scenarios include unmet demand are backordered, bulk demands and returns depending on the demands. The second model analyzes a model of a continuous-time production-inventory system with a shared server serving two demands. Unsatisfied demands are backordered and standard holding and penalty costs apply at each inventory location. The objective is to determine the optimal scheduling policy that minimizes the discounted total cost. The distinguishing feature of our model is that the two demand classes differ in their variability characteristics. The motivation for analyzing such a system is driven by our observations in many real-life contexts where such a difference is reported to exist. Very briefly, practical examples of why the variability of two demand processes, served by the same capacity, may differ include (1) inherent differences in the predictability of the demand, (2) some customers' spiky purchasing pattern influenced by low prices during promotions, (3) mixing emergency and scheduled demands, and (4) the nature of two different markets. Our results allow us to shed some light on implications for managers in such circumstances.