| This research introduces a novel approach for a production control policy. It is based on optimal control theory, which is utilized to regulate work-in-process (WIP), meet production target, and reduce production variability by on-line adjustment of the production rate of each machine in a manufacturing system. In this approach the buffer levels are the state variables, and the machines' production rates are the control variables. The control time interval is divided into sub-intervals dictated by the instantaneous operation states of each of the machines (working or needing repair). A library of optimal controllers is developed off-line for the sub-intervals, and the appropriate controllers are called from the library and applied to the manufacturing system according to instantaneous machine states.; The new control policy was evaluated through simulations under various conditions for two types of systems: serial manufacturing lines and reconfigurable manufacturing systems. The simulation results show that the proposed policy can successfully maintain low WIP and fulfill demands that are not very close to system's production capacity. Furthermore, the proposed policy is compared with the optimal policy, which is obtained by stochastic dynamic programming, for an example of a two-machine-one-buffer line. The results show that the proposed policy is able to achieve near-optimal performance with negligible differences for the specific objectives. Compared with dynamic programming (DP), which requires tremendous computation effort, this proposed policy can be obtained in only a small fraction of the time needed by DP, especially for a large system.; A heuristic rule for determining the policy's parameters is also proposed, such that the policy is suitable for various systems and various demands. Simulation results demonstrate that the new policy can be easily utilized to balance between maintaining low WIP inventory and reducing production variability, while the required production demand is still fulfilled. In general, the proposed policy can be adjusted for different requirements of WIP, throughput, and production variability, and is capable of providing feasible solutions for large manufacturing systems---a goal that is very hard to achieve with current known methods. |
