Research On One-of-a-Kind Production(OKP)Scheduling Problem

With the rapid development of social technology, customers personalized demand makes market environment change greatly. One-of-a-kind production(OKP) mode has become main production mode of most enterprises. Compared with mass production, One-of-a-kind production has more constraints, variables and calculation difficulties. OKP is becoming the focus of study.This paper begins with the background of the enterprise and focuses on the study of hybrid flow shop scheduling problem which is one style of one-of-a-kind production scheduling problem.Firstly, this paper considers general hybrid flow shop scheduling problem, constructs mathematical model and sets the optimal object as minimize makespan. This paper uses matrix to code and solves the problems by PSO. In order to avoid falling into local optimum, a SA selection mechanism is incorporated into PSO. The experiment on benchmark problems shows that the performance of improved algorithm is much better than general PSO.Secondly, this paper considers general hybrid flow shop scheduling model and optimal object with largest continuous processing time constraint, uses vector to code and maximum time remaining rule to choose machine. The solving algorithm is PSO. The experiment on benchmark problems shows that the results is optimal and the algorithm has better convergence.Thirdly, this paper considers general hybrid flow shop scheduling model and optimal object with batch constraint, uses matrix to code, sets the waiting time to determine the same batch size. The processing time of the same batch is equal to the maximum processing time of single job in this batch. The solving algorithm is PSO. The result proves the effectiveness of this algorithm.Finally, this paper considers the actual processing environment. With the example of S1line, this paper studies the hybrid flow shop scheduling problem with the constraints of maximum continuous processing time and batch process. The experiment is based on the actual processing data of S1. The result is better than original and the algorithm has good convergence.