Dynamic Scheduling Of Multi-objective Job Shops Based On GA-BPNN

The Job Shop Scheduling Problem(JSP)has been a very important research object in the field of production scheduling due to its complexity and practicality,and has received extensive attention from scholars at home and abroad.With the rise of Industry4.0,manufacturing enterprises are gradually moving towards diversified,customized and small batch production modes,which leads to more frequent occurrence of dynamic events in the machining process.Job shop scheduling problems with dynamic event disruptions offer greater flexibility,and optimising dynamic scheduling solutions can significantly improve a company’s productivity.In this context,it is all the more important to improve scheduling techniques in order to adapt to the changing needs of the market.As a result,speeding up the adjustment of production schedules to improve sensitivity and responsiveness to dynamic events has become an important challenge for manufacturing companies.To this end,this paper investigates multi-objective job shop dynamic scheduling optimisation methods around three dynamic events: machine breakdown,urgent order insertion and order addition.Firstly,this paper establishes a multi-objective job shop dynamic scheduling model with maximum completion time and total delay minimisation as the optimisation objectives and the machining sequence of workpieces on the machine as the variables,and establishes a corresponding rescheduling mechanism for the three dynamic events that occur more frequently in the manufacturing system.Then,in order to solve the multi-objective job shop dynamic scheduling problem,a multi-objective job shop dynamic scheduling algorithm based on GA-BPNN is proposed.It mainly includes:(1)the design of the network structure of the BP neural network according to the characteristics of the job shop scheduling problem;(2)the use of the CPLEX solver to solve the established mathematical model,thus providing training samples for GA-BPNN training;(3)the construction of training samples of the neural network for each of the three dynamic events;(4)the verification of the algorithm by arithmetic examples compared with the BP neural network and NSGA-II.Finally,using the actual scheduling problem as a case study,the initial optimal scheduling scheme of the case is obtained by using the CPLEX solver and visualized by relying on the simulation software Flex Sim;the occurrence of different dynamic events in the actual scheduling process is simulated and the trained GA-BPNN is used to rescheduling according to the rescheduling mechanism,and the scheduling results are compared with the heuristic algorithm NSGA-Ⅱ and the four commonly used scheduling rules The results are compared with the heuristic algorithm NSGA-II and four common scheduling rules to verify the effectiveness of the dynamic scheduling algorithm proposed in this paper,which can better and faster respond to the occurrence of dynamic events in real production processing.