Generally configured supply chain with flexible local control strategies

We consider a supply chain that comprises manufacturing stations, distribution stations and retail stations. A manufacturing station consists of a limited input buffer, a machine, a limited output buffer, a set of kanbans and a vehicle. A distribution or retail station consists of a limited storage buffer and a vehicle. A manufacturing station can choose any one of two mechanisms, buffer and kanban, to control its work-in-process, whereas that in a distribution or retail station is controlled only by a buffer mechanism. A vehicle can choose either a push or pull policy for transporting items from other stations to its own station. The supply chain is generally configured, which means that a vehicle can go to more than one station for transporting items rather than to a single upstream station. A manufacturing station can be controlled under a flexible strategy such as a push policy and buffer mechanism (push/buffer), or a push/kanban, or a pull/buffer or a pull/kanban, whereas a distribution or retail station can be under a push/buffer or a pull/buffer. Then, a global control strategy of the supply chain is formed by combining all local control styles. Such a supply chain is modelled as a queueing network. A decomposition approach is developed for major performance measures such as the throughput of each station and the work-in-process in each station. The efficiency of the decomposition approach is evaluated by comparing numerical experiments with simulations.