Research On Intelligent Workshop Scheduling Based On Deep Reinforcement Learning

The intelligent factory is an important carrier to realize intelligent manufacturing,and production scheduling is the key to ensuring the efficient operation of the intelligent factory.The job shop in intelligent factory senses and collects a large amount of production process data through technologies such as sensors and manufacturing Io T.By analyzing and training those real-time data of the job shop status,hidden scheduling rules and knowledge can be mined,and thereby optimizing the scheduling plan and realizing dynamic scheduling.Deep reinforcement learning can obtain scheduling strategies based on real-time production data,and optimize the selection of strategies according to the dynamic changes of the system state.In this paper,deep reinforcement learning is applied to job shop production scheduling in intelligent factory,and the following research is mainly carried out.(1)Scheduling environment and problem analysis of job shop in intelligent factory.Use the ISO/IEC 62264 standard under the industry 4.0 environment as a reference to analyze the types and processes of job shop scheduling in intelligent factory.The job shop scheduling problem in intelligent factory is described mathematically,and the new problems and requirements such as data-driven,real-time response,and intelligent decision-making of production scheduling under the intelligent factory are analyzed.(2)Job shop scheduling based on deep reinforcement learning in intelligent factory.In response to the requirements of data-driven in intelligent factory production scheduling,a scheduling method combining compound scheduling rules and deep reinforcement learning is established.With the goal of minimizing the maximum completion time,average flow time,maximum delay time,and maximum average delay,a conventional scheduling model is established to optimize the compound scheduling rules of rule conflicts and workpiece delays.Deep reinforcement learning is used to train production data and state parameters.On this basis,the algorithm automatically explores and selects the best scheduling rules at different decision points.(3)Multi-objective dynamic scheduling of job shop based on hypervolume in intelligent factory.Based on compound scheduling rules and Deep reinforcement learning scheduling.In view of the real-time changings in system data and dynamic scheduling optimization goals under the intelligent factory,in order to ensure the continuity of production in the event of dynamic uncertain events and improve the response ability to random events such as emergency order insertion,this paper combines hypervolume multi-objective optimization and action selection strategy to perform Pareto optimization of the algorithm for multiple objectives.(4)Empirical Research.In order to verify the application effects and prospects of the scheduling method based on deep reinforcement learning in the actual manufacturing environment,a machine tool manufacturer was used as the object to analyze the actual production scheduling environment,workpiece processes,equipment conditions and parameters of the factory.New orders are generated in the same batch,and the scheduling plan is solved based on deep reinforcement learning.Also,a dynamic scheduling plan is solved for the urgent orders that arrive randomly.The research results show that deep reinforcement learning can make production scheduling decisions through data training based on real-time changes in system status,and effectively realizing autonomous and dynamic production scheduling in the job shop under intelligent factory.