A genetic algorithm based knowledge acquisition system for scheduling FMS

Dynamic selection of scheduling rules during real time operations has been recognised as a promising approach to the scheduling of flexible manufacturing systems(FMSs), which are designed to handle a large and constantly changing variety of parts. This paper sets out an approach which extracts scheduling knowledge to suggest a set of priority dispatching rules(pdrs) one each for one workcell(WC), which are to be followed during different planning horizons, based on the WC process attributes for the scheduling problems in FMS. The proposed methodology includes three modules: training example generation(TEG), knowledge acquisition and learning(KAL), and scheduling rule generation(SRG). in the TEG module, for the part mix data generated randomly, the WC attributes are extracted and classified, and the optimal WC-wise pdr set is obtained through a genetic search process. Then, they are combined together as one set of training examples. The KAL module establishes the relationship between the WC attribute classes and its respective optimal pdr with a binary tree structure. With many different problems of various sizes generated with TEG module and the MAL module, a pdr which dominates the others in each and every combination of WC attribute classes is singled out and formulated as scheduling knowledge base rules in SRG module.