APPLYING ARTIFICIAL-INTELLIGENCE TECHNIQUES TO MODELING MANUFACTURING SYSTEMS

Artificial intelligence techniques promise significant advantages in discrete simulation modelling. In particular object-oriented programming offers the possibility of mapping simulated entities onto programme objects, so that class inheritance simplifies definition of similar, non-identical entities. A generalised object-oriented discrete simulation environment recently released as a commercial product provides an efficient model-building interface, comprehensive analysis facilities, and the capability of observing and modifying model behaviour during execution. However more important benefits are firstly that by specialising standard objects any desired complexity can be added to the environment, and secondly that a complex subsystem model can be defined as a composite object and included in higher level simulations. Such composites may also be replicated, either exactly or with some modification in the same or other simulations. This latter facility also offers the possibility of replacement of a subsystem model by a representative metamodel, where the performance of a composite representing part of a modelled system may be analysed so that a statistical model object can replace the composite, thus enhancing model performance. The commercial product is designed for general application, so standard object classes are designed to supply a generic range of functionality, and are named to reflect their function. The intention is that the process of specialisation can be used to modify the behaviour and nomenclature of standard classes to suit particular application needs, whilst more complex functionality, normally only used recurrently within an application area, can be provided by creation of composite object classes. Libraries containing both specialisations and composites are then constructed. Such libraries can be loaded into the programming environment to facilitate modelling in the specific application area. The methods of communication between objects defined within the commercial product are sufficiently general to operate without modification, and thus provide a protocol for transmission of objects specialised to meet application needs. This paper describes the development of a specialised manufacturing 'personality' of the environment including a range of typical manufacturing and control system objects, from the level of detail of individual machines up to generalised definitions of assembly lines and production control functions. The development has allowed modelling of real, complex manufacturing systems, and forms the basis for discussion of the benefits and difficulties of simulation in an object-oriented environment.