A simulation and learning technique for generating knowledge about manufacturing systems behavior

One of the most important difficulties when developing knowledge based systems in manufacturing scheduling or control, is finding the required knowledge. The authors address here the problem of acquiring knowledge about the behavior of manufacturing systems. Learning algorithms are proposed to generate, from simulation experiments, a set of production rules. This set may be considered as a simulation meta-model, and may be used either directly by the shop manager, or inserted into a knowledge base. This approach is illustrated by the use of the learning program GENREG. It generates rules related to the behavior of a simplified flow shop when different dispatching rules are applied.