Study On Multi-strategy Dynamic Scheduling Optimization Algorithm In Rotating Seru System

Japan sera system is a combination of flexible job shop,mass production mode and the sustainable development of environment friendly etc,and is widely used in the United States,Europe,South Korea and China and other countries of the electronics industry.It is a new kind of production management mode of human-centered and low automation.There are three main types of seru production systems:division,rotating and yatai.Since its emergence,the seru system has received wide attention from the business communities and the academic circles for its efficiency and adaptability to many varieties,fewer batches and more batches of market demands.The scholars focus on two problems of seru production system:seru formation and seru loading.Over the past two decades,there has been a lot of academic researches on seru formation,and there is relatively little literature on seru loading.The seru loading problem consists of two parts:the matching relationship between the product group and the seru,and the scheduling of employees and the jobs in the seru.This paper mainly focuses on the scheduling of employees and jobs in the seru.In the rotating seru production system,in the face of dynamic market demand,three are new jobs arrived.We use multiple strategies of memetic algorithm to solve the scheduling problem,where minimized the total flow time(TFT)of jobs and total labor hours(TLH)for employees for the bi-objectives.Memetic algorithm combines the global search and local search.The idea is to simulate the effort of the majority of the general public and a few social elites with professional knowledge in order to promote the progress of social culture.The memetic algorithm of this paper uses NSGA-II as global search algorithm,and the two improved NEH algorithms as local search algorithms.In order to improve the convergence speed and search ability of the algorithm,four optimization strategies are designed for the characteristics of the rotating seru system:(1)NEH.SPT initialization population strategy.For the Japanese seru production system,seru is converted from assembly line,retaining a certain flow shop property.And NEH is one of the best heuristic rules for flow shop.Therefore,this paper optimizes the initial population using NEH and SPT rules.(2)Graph theory initial population strategy.For optimizing the total labor hour,,there is a one-to-one relationship between employees and job sets in the rotating seru production system.By transforming the corresponding relationship between employees and job sets into assignment problem,we can use graph theory to solve the minimum cost maximum flow theory in to make the minimizing total labor hour.Because the work order of the staff is consistent on all rotates,we can see the scheduling of employees and their corresponding job sets as the scheduling problem of the jobs with uncertain processing time.In this paper,with an average processing time of job sets,the mathematical model is set up to solve the scheduling of the job sets that is the employee's work order.Finally,the NEH_S algorithm based on the idea of NEH is used to optimize an individual with a better bi-objectives value,which is going to add into the initial population to improve the convergence speed and search quality of the algorithm.(3)two improved NEH rules for local search.In rotating seru system,employee finish all the jobs for a number of rotates,at every rotate,jobs operations can be regarded as a small flow shop problem,for each circuit this paper uses NEH rules-NEH_D algorithm based on decomposition optimization method.The NEH_D algorithm and the NEH_S algorithm perform local search operation on the individual with the same probability.(4)dynamic prediction scheduling.In multiple dynamic scheduling,it can be believed that each scheduling result contains a number of better working sequence chromosome.So this population will take some actions like particle swarm optimization(PSO)algorithm.The step length is the Euclidean distance between the two centers of the pareto front of first and second dynamic scheduling,and the direction is direction of the center of the individuals with the worse object value and the center point of the individual with the better object value.We will improve the individuals with the poor object value.In order to verify the search performance of the improved memetic algorithm,this paper uses NSGA-?,HQGA,MOEAD algorithm as the comparison algorithm to evaluate the IGD and HV as the evaluation metrics,and the classic examples of Taillard is used for the test.According to the results of numerical experiments,in different dynamic largest quantity of jobs and different number of initial jobs,IGD and HV of the improved memetic algorithm indicators are better than the other three algorithms,and along with the increase in the quantity of jobs,the memetic algorithm is more obvious advantage.When evaluating the four strategies separately,it was found that the performance of individual local search strategy gradually decreased with the increase of the number of jobs,and the performance of the initial population strategy of the NEH and SPT is gradually enhanced.The performance may be caused by multiple dynamic scheduling.