Processing Networks with Inventories: Sequential Refinement Systems

This paper investigates a model representing a make-to-stock production system, where intermediate and finished goods are produced and stored in advance of demand. In other respects the system operates much like a network of queues. There are several nodes and items. Each item is produced at one of the nodes using another item as input. Customer demands for the items and the unit-processing times at the nodes obey simple Markovian laws. We develop a tractable approximation scheme for estimating performance and test it through computer simulation. Numerical results show that the approximation is quite accurate. The model and the approximation express the effects of different degrees of component commonality.