Research On Dynamic Job Shop Scheduling Method Based On Dispatching Rules

Workshop scheduling is the core link in the control,optimization and management of the entire production workshop.The effective scheduling algorithm based on dynamic job shop can reduce manufacturing costs,shorten product delivery time,improve customer service quality and improve the overall performance of the manufacturing system.This thesis takes the dynamic job shop scheduling problem as the research object,and proposes a dynamic job shop scheduling method based on machine learning and dispatching rules.A overview of the domestic and foreign research status of scheduling methods based on dispatching rules is carried out,the deficiencies of heuristic methods,single rule and traditional intelligent optimization methods in dealing with dynamic scheduling problems are analyzed,a dynamic scheduling algorithm based on Q learning algorithm and dispatching rules to solve dynamic scheduling problems is proposed.“Minimize tradiness time” is selected as the performance indicator,a state space of the production process centered on “the urgency of remaining tasks” and a reward function with the purpose of “the higher the slack,the higher the penalty” are established;an action group containing 6 dispatching rules is formed;an action selection strategy based on the "Softmax" function is designed to replace the original “greedy strategy”,and the improvement of the Q learning algorithm is completed.Computer simulation is carried out in a dynamic environment where the work to be processed arrives at the workshop in an orderly manner,the results show that the tardiness time of the algorithm is much shorter than that of using a single rule and traditional intelligent optimization methods,which verifies the effectiveness and superiority of the improved Q-learning algorithm in solving the dynamic job shop scheduling problems.Aiming at the problem of complex job shop dynamic scheduling with the nature of “highdimensional input data”,a dynamic scheduling algorithm based on DQN and dispatching rules is proposed.“Minimize tradiness time” is selected as the performance indicator,5 state characteristic equations that can more comprehensively describe the production state are put forward,and a mathematical expression of return function;an action group containing 10 dispatchling rules and a better learning “Softmax” function as an action selection strategy are designed and formed respectively;the activation function used in the training of the deep neural network is determined,and finally the construction of the overall framework of the neural network in the DQN algorithm is completed.The simulation results show that the algorithm can select appropriate dispatching rules according to different scheduling moments to reduce the tradiness time.At the same time,the algorithm has good convergence under the set conditions,and the results obtained are far better than using a single scheduling rule,Q learning algorithm and traditional intelligent optimization methods,which verifies the effectiveness and superiority of the scheduling algorithm in solving the dynamic job shop scheduling problems.