APPLICATION OF IMPROVED HOPFIELD NETWORK TO CYCLIC FMS SCHEDULING

This paper proposes a model of an application of Hopfield neural network to the Flexible Manufacturing Systems scheduling problem. The cyclic scheduling of this problem is NP-hard. This model describes details in calculating Work In Progress from the schedule taking into account the corresponding linked, precedence and disjunctive constraints. An unconstrained optimisation model is formulated and a methodology of applying Hopfield neural network is given. Three different scheduling benchmarks are used for testing and comparative experimental results are provided, with a conclusion discussed indicating the advantage of this approach.