Rule-based simulation metamodels

Simulation models provide useful knowledge about the dynamic behavior of systems. It has been proposed to incorporate these models in knowledge-based systems. Unfortunately, simulation is time and memory consuming and the results, which are frequently in the form of a variety of large tables of data, are often difficult to interpret. Due to these problems, simulation metamodels (models of simulation models), such as regression metamodels, have already been suggested. The paper introduces rule-based metamodels, which are a set of production rules that model the simulation model. They can be obtained from simulation results using automatic learning programs. The rules represent understandable knowledge that can be directly used for decision making, or inserted into a knowledge-based system. The present approach is illustrated and discussed through manufacturing scheduling and control problems. It is also suitable to other Artificial Intelligence formalisms, such as frames.