A Study On Apparel Sewing Production Scheduling Method Based On Deep Reinforcement Learning

With the development of Industry 4.0 and "Made in China 2025" strategy,the issue of production line arrangement and scheduling in garment production companies has attracted more and more attention,which is a key problem to improve production efficiency and core competitiveness of garment enterprises.Production line arrangement and scheduling refers to the rational arrangement of equipment and garment production sequence at workstations on the production line under current production resource conditions to improve production efficiency.In this paper,taking the sewing production line of a garment production enterprise in Henan as the research object,we model and solve the optimization problem of multi-objective sewing production line workstation arrangement and the real-time dynamic scheduling problem of sewing process,and develop a set of garment sewing production line arrangement and scheduling system.By using the system to provide decision support for garment sewing production,the time required for determining the arrangement and scheduling plan of the sewing production line is reduced,and the production efficiency of the enterprise is improved.Firstly,the garment sewing workshop of a garment production enterprise in Henan was investigated,and the development goal of the system was determined and the production information database was established by summarizing the production demand and efficiency bottleneck of the enterprise.Secondly,aiming at the current lack of research on the optimization problem of multi-objective sewing production line workstation arrangement,a sewing production line workstation arrangement optimization problem model was established with the goals of sewing production line balance rate and equipment loss.The multi-objective optimization algorithm NSGA-II was used to solve the model,and the results were compared with manually arranged results.The arrangement plan provided by this method effectively improved the sewing production line balance rate,reduced equipment loss,and greatly reduced the time required to generate arrangement plans.Thirdly,a garment sewing process scheduling optimization model based on deep reinforcement learning was proposed to minimize the maximum completion time,which is a common problem caused by the dynamic factors in the sewing process that affect the effectiveness of static scheduling.This model was transformed into a sequential decision-making problem based on Markov Decision Processes.State features,candidate action sets,reward functions,exploration and exploitation strategies,and deep neural networks were defined to describe state-action values,and the most suitable scheduling rules were selected at decision nodes.The experimental results showed that,compared with the genetic algorithm,the proposed method was equivalent in achieving scheduling goals but significantly improved decision efficiency.Furthermore,it could respond in real-time to the problem of dynamically arriving orders,ensuring the effectiveness and continuity of the scheduling plan.Finally,based on the above research,the garment sewing production line arrangement and scheduling system was implemented using Python as the development language and My SQL as the database management system.The system has achieved expected functions and can provide decision support for garment production enterprises.