An intelligent controller for manufacturing cells

To meet multiple performance objectives and handle uncertainty during production, a flexible scheduling system is essential. In this study, a neural network based control system is proposed to adapt different scheduling strategies dynamically for a manufacturing cell. The proposed control system consists of an adjustment module and the associated equipment controller for each machine and the robot. At a decision point, the adjustment module will determine the relative importance for each performance measure according to the current performance levels and requirements. The equipment level controller, implemented by a neural network, will select the proper dispatching rule based on its status and the relative importance levels. The problem, which arises from the discrepancy of the user specification and what neural networks are trained by, is addressed. From the simulation results, the proposed refinement procedure could recover this problem so that the controller can perform closer to the actual requirements. In addition, the performance of the controller in the multiple criterion environments and its adaptability are investigated through simulation studies. The results show that this proposed controller performs well under the multiple criterion environments and is able to respond to changes in objectives during production.