Multi Resource Scheduling Of Discrete Manufacturing Workshop Considering Material Handling

With the complex change of product demand in discrete manufacturing workshop and the improvement of intelligent and digital level of workshop equipment,the production process is changeable,the types of workpieces are increased,and the process route of workpieces is more flexible.A reasonable workpiece scheduling scheme will help to improve the production capacity of the workshop.With the development of intelligent logistics system,the workshop needs AGV trolley to shuttle between different areas to complete the transportation of workpieces between equipment and ensure that workpiece processing can be carried out according to the scheduling scheme.AGV scheduling scheme and path selection will affect the execution of workpiece processing and scheduling.Therefore,the reasonable scheduling decision of multi resources such as processing equipment and handling trolley in discrete manufacturing workshop has certain theoretical significance and engineering application value for improving the implementation rate of production scheduling scheme in manufacturing workshop,ensuring product delivery date and improving the overall performance of manufacturing system.The subject mainly focuses on the multi resource scheduling problem of discrete manufacturing workshop considering material handling.The research contents are as follows:Firstly,the production process and production characteristics of discrete manufacturing workshop are analyzed,and the multi resource scheduling problem of discrete manufacturing workshop considering material handling is decomposed into multi resource（processing equipment and AGV）scheduling problem and trolley path planning problem between equipment,so as to construct the mathematical model of multi resource scheduling problem and reinforcement learning environment.Aiming at the maximum completion time of workpiece and the balance of AGV load,considering the constraints of workpiece process route and machining characteristics,a multi resource scheduling mathematical model of discrete manufacturing workshop is established.According to the relevant parameters involved in the mathematical model of multi resource scheduling,the state space of reinforcement learning is designed,the action space of reinforcement learning is designed through the executable action of multi resource scheduling,and the reward function of reinforcement learning is designed according to the goal of the mathematical model.Taking the shortest path,the least number of turns and no collision between AGV as the goal,and considering the constraints such as workshop layout,the mathematical model of multi AGV path planning in discrete manufacturing workshop is established.The reinforcement learning state space is designed according to the grid model,the reinforcement learning action space is designed through the AGV action in the workshop,and the reward function of reinforcement learning is designed according to the objectives in the multi AGV mathematical model,which provides the basis for the subsequent multi resource scheduling algorithm and path planning algorithm.The multi resource scheduling method of discrete manufacturing workshop based on DQN（deep reinforcement learning）algorithm is studied.Based on the multi resource scheduling process,the learning process of multi resource scheduling algorithm extracting scheduling information from experience pool and minimizing output loss is analyzed,and the job scheduling agent and AGV scheduling agent are designed to realize the multi resource scheduling of workshop.The mechanism of scheduling agent is studied,including the local state space,action space and local reward mechanism of job and AGV scheduling agent based on the scheduling function and mathematical model objectives.Through neural network the decision-making process map the relevant information of job and AGV scheduling into weight information and synthesizing the scheduling list.Design the ε-greedy action selection mechanism with the decision-making process of scheduling list.interaction mechanism between job and AGV agent,multi resource scheduling algorithm and multi AGV path planning algorithm is designed.Through the actual scheduling case in discrete manufacturing workshop,the effectiveness of this algorithm is verified.The multi AGV path planning method based on improved Q-learning algorithm is studied,and the multi AGV path planning process is analyzed.The multi AGV path planning method is studied,including based on the artificial scalar field,combined with the location of obstacles in the workshop,the starting point and ending point of the car design the Q table initialization method.Based on ε-greedy and Boltzmann methods,the exploration and utilization strategy is improved by combining the current iteration times of the algorithm,the algorithm iteration formula is improved by adding a learning layer,and the dynamic learning rate is designed by combining the exploration and utilization strategy and two-layer learning layer.The effectiveness of the algorithm is verified by comparing the algorithm with the path planning test case.Finally,combined with the actual production requirements of discrete manufacturing workshop,a prototype system of multi resource scheduling in discrete manufacturing workshop is developed.The overall framework of the prototype system,each functional module and the database related to multi resource scheduling are designed.Python,Py Qt and SQLite databases are used to develop the prototype system,including the function realization of login module,basic information management module and multi resource scheduling module,so as to help the production scheduling planners in discrete manufacturing workshop make decisions.