Prof. Reha Uzsoy (North Carolina State University): Fast Approximate Solutions for Capacitated Production Planning Under Uncertain Demand – The effective management of supply chains producing physical goods requires coordination of the production systems manufacturing the goods, the inventories held for different reasons throughout the supply chain and the logistics systems that move the goods from origin to destination. Despite the evident interdependencies, however, the domains of production planning and inventory management have diverged widely over recent decades. Production planning has tended to focus on deterministic optimization models, emphasizing the representation of complex technological and material constraints. Inventory management, on the other hand, has focused on modelling the stochastic nature of demand, with less interest in how inventories are replenished.
This talk will present a body of work that seeks to link the two areas of inventory management and production planning. Since the underlying stochastic optimization problem is complex, rendering exact solutions impractical, we seek to obtain approximate solutions using simple linear programming models that i) account for queueing behavior in the production system, specifically workload-dependent lead times, and, ii) use chance constraints to maintain appropriate safety stocks in the face of stochastic demand. Recognizing that most production systems are driven by forecasts that are updated as information evolves, we use the Martingale Model of Forecast Evaluation as a demand model to derive the chance constraints. Extensive computational experiments in several different environments show promising results and indicate directions for further improvements.
