Research On Distributed And Flexible Job-shop Scheduling Algorithm Based On Deep Neural Network

In recent years,with the globalization of the economy,the organization of industrial production has been largely changed by the distribution of transnational and cross-regional companies and the increasing cooperation of small and medium enterprises.At the same time,in the increasingly fierce competition in the global market,users’ personalized customization demands make the customer-driven,one-of-a-kind production mode become the trend of the modern manufacturing industry.Under such a background,distributed and flexible manufacturing emerges as the times require.Distributed and flexible manufacturing has the features of high flexibility,high dynamic,high agility,and geographical distribution of manufacturing resources,which greatly enhance the complexity of scheduling under such environment and make the traditional scheduling algorithms cannot solve this kind of problem well.With the development of machine learning related technology,the data-driven intelligent scheduling algorithm has brought a new opportunity to solve the problem of complex production scheduling.Inspired by the idea of “ Data + Learning”,starting from data-driven intelligence,the intelligent algorithms for distributed and flexible job-shop scheduling problem based on deep neural network is studied.The main works of this paper are summarized as follows:First of all,the distributed and flexible job-shop scheduling problem under the modern one-of-a-kind production mode is analyzed.According to the three-stage scheduling structure of distributed and flexible manufacturing,the three-stage scheduling problems,which are job scheduling,operation scheduling and operation sequencing,are modeled respectively.Focusing on the constraint of orders’ delivery deadline,the global scheduling optimization objective is to minimize the average tardiness penalty.And according to the scheduling tasks of each scheduling stage,the sub optimization objective is set up for each scheduling stage.Then,after fully considering the dynamics of distributed and flexible job-shop scheduling and analyzing the diversity of scheduling influencing factors,an intelligent scheduling algorithm based on deep neural network,for solving distributed and flexible job-shop scheduling problem,is proposed.Aiming at solving the resources assignment problems of job scheduling and operation scheduling stages,The algorithm uses massive,multi-dimensional simulation scheduling historical data to train deep neural network.After that,it can effectively evaluate available resources according to the realtime tasks’ and resources’ related data,and select the best resources for jobs and operations.While,for solving sequencing problem on a specific machine,the scheduling model,which is based on the deep neural network,learns from the simulation scheduling historical data and it can reasonably allocate the processing priority for operations.Finally,the scheduling algorithm can realize the optimization of the scheduling objective.At last,using Java and Python,with the help of related tools,software simulation is carried out.The effectiveness of the proposed algorithm is verified by contrast experiments of simulation scheduling.In addition,a multi-agent-based intelligent distributed and flexible job-shop scheduling system framework is designed.Combined with modern cyber-physical system technology,the feasibility of the proposed scheduling algorithm in the actual production process is expounded.This research provides a new idea to solve the distributed and flexible job-shop scheduling problem under the modern one-of-a-kind production mode,and it is of great significance for the promotion of both enterprises’ benefits and users’ satisfactions.