Setup time reduction in linear production systems.

摘要

Current industrial experience suggests that reducing setup times in a stochastic multiple product flow shop with finite buffers increases the capacity of the shop. Prior research on setup time reduction has not examined the structure of this relationship in shops with several workstations and finite work-in-process inventory buffers. One empirical and five analytical models estimating the additional capacity to be gained from reducing setup times are developed. An empirical Response Surface Model is fitted to data generated from a simulation model of a stochastic, multiple product, finite-buffered flow shop using a sequential experimental strategy. The Response Surface Model suggests that reducing setup times on the non-bottleneck machines becomes important as the magnitude of their setup times approaches that of the bottleneck machine; in addition, setup time reduction is most beneficial when the individual product lot sizes are small. The five analytical models are comprised of three bottleneck machine models and two multiple machine models. The bottleneck machine models approximate the capacity of the shop as that of the bottleneck machine, while the multiple machine models consider the interaction between adjoining machines in estimating shop capacity. Of the three bottleneck models, the first is deterministic, the second assumes the bottleneck service process to be Poisson, and the third permits the bottleneck service process to be any general renewal process. The first multiple machine model assumes that the service processes at each workstation are Poisson, while the second models the service processes as general renewal processes. The accuracy and bias of each of the analytical models are then examined. The Poisson Bottleneck Model is slightly more accurate than the General Multiple machine Model, but the General Multiple machine Model is by far the least biased of the five analytical models. The General Multiple machine Model therefore appears to be the preferred analytical model for analyzing setup time reduction.