A neural network model for on-line control of flexible manufacturing systems

This research investigates the potential of using a neural network approach in real-time control of flexible manufacturing systems. A hierarchical manufacturing controller, consisting of two neural network structures, is proposed. The first neural system participates in the feasibility analysis, and the other, at the lower level, in the process of dispatching and control. At the first level, a Sigma-Pi type of connection is used to translate work-in-process (WIP) move requests into directed arcs. Through a filter scheme, infeasible arcs are identified and eliminated from further consideration. At the second level, a modified Hopfield-Tank model is developed to determine the correct moves. Its goal is to deliver the right WIP to the right workstation, and process it at the right time. An example is used throughout the paper to illustrate the architecture developed. This two-phase control procedure provides adaptability, speed, and good solution quality which are important for real-time control of flexible manufacturing systems.