The control of a flexible manufacturing system by short-term goal identification

The control of manufacturing systems is characterised by the need to make a compromise between conflicting goals. Traditional research concentrates on steady-state optimisation of single objectives that are not necessarily relevant to decision making in a production environment. The advent of flexible manufacturing systems (FMSs) with automated material handling and limited buffer space highlights the need to provide computationally-efficient solutions to work out scheduling and dispatch problems for real-time applications. The authors have adopted a knowledge-based approach to achieve effective real-time FMS control. In the context of the current system status and planned system activities, short-term goals are identified which lead to appropriate jobdispatch decisions. This approach has been implemented in a prototype control system written in Prolog. To obtain knowledge in this area of scarce expertise, a “define-build-learn” cycle was adopted. First, functional concepts are defined and built into the control system, then the corresponding system characteristics are studied and the results used for further cycles of concept definition and building. This paper describes the development of this operational system.