IMPLICIT VERSUS EXPLICIT KNOWLEDGE REPRESENTATION IN A JOB-SHOP SCHEDULING DECISION-SUPPORT SYSTEM

Job-shop scheduling is a key problem for many manufacturing firms. Due to the underlying complexity, only heuristic methods can be used to s olve this problem. The choice of a relevant scheduling strategy is bas ed on the workshop and the manufacturing orders structure and on the o bjectives of the workshop manager. We show in this paper how to take t his information into account in order to select a relevant scheduling strategy. A first approach is to represent the expertise on this domai n explicitly. The managed information, such as the importance of an ob jective according to the workshop manager, is often imprecise. Therefo re a fuzzy inference system is described to support knowledge represen tation. The second approach is to learn and to generalize the expertis e implicitly from several typical situations thanks to neural networks . Both approaches are described and compared with an industrial schedu ling software.