Performance evaluation of closed tree-structured assembly systems

Abstract

Assembly systems are very common in manufacturing environments. In such systems, certain machines require more than one part type before they can start processing. Our paper presents a practical stochastic model for evaluating the performance of closed-loop assembly systems. This evaluation is difficult because of synchronization delays and the large state space required to capture the interaction of the flow of parts among the various subassemblies. We present an approximate, but computationally efficient, aggregation/disaggregation algorithm to compute the system throughput and mean queue lengths for both the matched and unmatched WIP inventory components. The aggregation computes the conditional system throughput as a function of the number of jobs in the system, whereas the mean queue lengths are obtained by disaggregation. Application of our algorithm reveals several salient operational insights, such as the increase in queue length towards the downstream machines or the concavity of the assembly system throughput with respect to the total number of kanbans. © 1996 "IIE".