Research On Complex Network Features Based Neural Network Scheduler For Job Shop Scheduling Problem

The research of production scheduling plays an important role in improving the production efficiency of enterprises and reducing the loss of resources.Although the researchers have made a lot of research work on solving scheduling problems and proposed many scheduling algorithms,most of them are difficult to meet the real-time requirements of the real-time production process due to the slow solution speed.The traditional scheduling rules are widely used in actual production processes due to their low computational complexity,easy implementation,and easy understanding.However,a single scheduling rule is often difficult to obtain a good scheduling solution,especially in a dynamic production environment.A single scheduling rule is difficult to adapt to a dynamically changing scheduling environment.Therefore,how to quickly obtain a better scheduling scheme in an acceptable time range and effectively guide the actual production work has become a hot topic in the research of scheduling problems.On the other hand,information technology has been used in manufacturing enterprises,and manufacturing companies have accumulated a large amount of valuable data related to scheduling,and data-driven scheduling methods have also been studied more and more.This paper studies how to use neural network scheduler to solve job shop scheduling problems.The research in this paper is as follows:(1)Aiming at the problem of static job shop scheduling,a dual BP neural network scheduler is proposed.The existing scheduling scheme is transformed into a training data set to train the neural network,and the generated neural network scheduler can quickly and effectively solve the new scheduling problem at the same scale.Experiments show that the generated scheduler solves faster than the existing intelligent algorithms,and the obtained scheduling scheme is superior to the traditional scheduling rules,and other current data-driven scheduling methods.(2)For the dynamic job shop scheduling problem,considering the dynamic arrival of the workpiece,a data-driven dynamic job shop neural network scheduler is proposed.The genetic algorithm is used to solve the static scheduling problem,and the better scheduling scheme is transformed into the training data set according to the corresponding scheduling attributes,and the neural network is trained.The obtained neural network scheduler can effectively sequence any two conflicting processes,so as to solve the scheduling problem that the workpiece dynamically reaches in the dynamic job shop scheduling problem.Experiments show that the proposed method can solve the dynamic job shop scheduling problem quickly,and the obtained scheduling scheme is superior to the traditional scheduling rules.(3)Aiming at the influence of scheduling attributes on the accuracy of neural network classification,thus affecting the performance of the final output scheduling scheme,a method of scheduling attribute extraction based on complex network modeling is proposed,which can systematically extract potential scheduling attributes.The scheduling properties complement the existing scheduling properties.Experiments show that the complex network features extracted according to the complex network model can effectively reflect some potential information of the scheduling scheme.These complex network features can effectively transform the scheduling scheme into training datasets to train the neural network scheduler and generate better scheduling.