Setting Basestock Levels in Multiproduct Systems with Setups and Random Yield

This paper shows how to set basestock levels in a multiproduct system with setups and random yield. The system is represented by a polling model and the inventory level of each product is controlled using a basestock policy. When the queue is empty, the inventory level is equal to the basestock level, thus the server will continue to serve the queue until it is empty. If the capacity of the queue is equal to the inventory allocated to the item, then when the queue is full, inventory is fully depleted and new demand is either backlogged, lost or expedited. Defects are routed to temporary storage queues associated with each item, and then routed back to the original queue for service during the next cycle. For a system with backlogging, we provide a cost function that is minimized by solving N single item newsvendor problems. For a system with lost sales or expediting, we introduce a cost function and provide a heuristic for finding the basestock levels. The effectiveness of the heuristic and accuracy of the cost approximation are validated through numerical tests.