A forecast-driven tactical planning model for a serial manufacturing system

We examine tactical planning for a serial manufacturing system that produces a product family with many process steps and low volumes. The system is subject to uncertainty in demand, in the supply of raw materials, and in yield at specific process steps. A multi-period forecast gets updated each period, and demand uncertainty is realised in terms of forecast errors. The objective of the system is to satisfy demand at a high service level with minimal operating costs. The primary means for handling the system uncertainty are inventory and production flexibility: each process step can work overtime. We model the trade-offs associated with these tactics, by building a dynamic programming model that allows us to optimise the placement of decoupling buffers across the line, as well as to determine the optimal policies for production smoothing and inventory replenishment. We test the model using both data from a real factory as well as hypothetical data. We find that the model results confirm our intuition as to how these tactics address the trade-offs; based on these tests, we develop a set of managerial insights on the application of these operating tactics. Moreover, we validate the model by comparing its outputs to that from a detailed factory simulation.