JOB-SHOP SCHEDULING DESIGN WITH ARTIFICIAL NEURAL NETWORKS

Abstract

Simulation, being capable of representing a system's behavior in an effective way, when combined with the neural networks, can provide an efficient decision making structure. In this paper, a system is developed in order to determine the machine, the material handling system and the priority rule that will be used in the system by using Simulation and neural network techniques in Job-Shop scheduling design. The backpropagation algorithm is chosen for the neural network model. In this paper, first, a neural network that is capable of providing realistic results is obtained. Simulation technique is used in order to obtain the samples to train the neural network in computer environment. The next step includes the decision-making, determination of the ranges where the selected decision remains valid and the related comments. Trained neural networks are used in order to determine the hardware configuration and the scheduling strategy that are capable of providing a determined set of performance criteria. After the simulation of the result(s) that is (are) proposed by the neural network, the deviations of the performance criteria from their corresponding expected values are calculated and proposed in a tabular format. The criteria used in the performance measurement are the average flow time, average tardiness, maximum completion time and machine center usage ratios.