A multi-agent simulation methodology for manufacturing systems

Performance evaluation of manufacturing systems supposes a description as precise as possible of its functioning before the elaboration of a simulation or analytical model. The modelling process is decomposed into two steps: the building of a knowledge model. (specification step) and its translation into an action model (evaluation step). The use of a method combined with a graphical tool in order to build the knowledge model permits a better validation of the study by experts. A multi-agent model is used to specify the decisional sub-system. We show that this subsystem is structured, hierarchical and complete: most of manufacturing problems can be solved. The multi-agent methodology is then used to dimension a manufacturing system which produces coils. This methodology was validated on other problems.