Neural network based optimization in production scheduling

Although a large number of approaches such as mathematical programming, dispatching rules, expert systems, and neighborhood search to the modeling and solution of scheduling problems have been reported in the literature, over the last decade, there has been an explosion of interest in using artificial neural networks (ANNs) for the solution of various scheduling problems. The objective of this research is to utilize ANNs to deal with two different scheduling problems. The first problem considered is the classical identical parallel machine scheduling problem with makespan minimization. A dynamical gradient type neural network, which employs a penalty function approach with time varying coefficients, is proposed for the solution of the problem. Simulation outcomes of the proposed approach are compared with those of the longest processing time rule and with the optimal solutions, for different sizes of scheduling problems. The second problem is the scheduling of an independent jobs set with sequence-dependent setups and distinct due dates on nonidentical multi-machines to minimize the total weighted earliness and tardiness. The original mixed integer formulation of the problem is modified by adding one more constraint to the model to prevent the assignment of two jobs at the same time to the first position. For this problem, an interconnected neural network model which is composed of two maximum, three piecewise linear and one log-sigmoid neural networks is proposed. The proposed approach is tested on a scheduling problem and the results are compared with optimum results of the linear programming solver.