Multiple machine feedforward networks with blocking
Conference Paper
Overview
Additional Document Info
View All
Overview
abstract
2018 Institute of Industrial Engineers (IIE). All rights reserved. Almost all of the analytical decomposition approaches for queueing networks with limited WIP allow only single machine workstations. This paper considers the development of a multiple-machines feedforward network analysis system for exponential servers and finite capacity workstations (Kanban control). This approach can handle nonidentical service rates for the different machines at the workstations. Multiple workstation systems quickly become too complex for a single steady-state model and decomposition into single workstation methods are utilized. These results are connected via adjusted flows and increased delays and require an iterative scheme that cycles through the individual approximation systems in a manner such that the overall throughput result converges. The sequencing issue of blocked jobs causes the complexity of the merge configuration to increase significantly as the number of merging machine flows increases. The aggregate state-space theorem is extended to the multiple machines/workstation case which allows a workstation sub-model with only one state for each level of the number of blocked jobs. Then these aggregate probabilities yield the sequence-depend probabilities needed to develop the appropriate blocking probabilities for each inflow machine. The approximation system performance measures agree quite well with simulation generated results.