Throughput and inventory estimation of a pull-based supply system

We analyse an assembly system fed by components produced by suppliers and operating under a pull system with random processing times. Under certain assumptions we present a performance evaluation algorithm which is based on physical decomposition of the network and use of an aggregation/disaggregation algorithm. We consider two important performance measures: system throughput and work-in-process inventory. Performance evaluation of these systems is difficult because of the mating of parts and a kanban-like operating mechanism. Our algorithm was tested using simulation and found to be accurate.