| With the improvement of economic level and life quality and the change of market environment and customer demand,the modern manufacturing mode has continuously evolved to a multi-variety,small-batch mixed-line production mode.More and more manufacturing enterprises adopt mixed-line production mode for production in flexible job shop.At the same time,after the "Made in China 2025" strategy was put forward,AI-based job shop scheduling methods have become a new research hotspot.Therefore,establishing a mathematical model for mixed-line production job shop scheduling problem and studying AI-based job shop scheduling methods are of great significance for guiding modern enterprises’ production.In this paper,Agent technology is combined with job shop scheduling technology and the mathematical model for this problem is researched.A multi-agent scheduling method for this problem is proposed and verified by experiments.The specific research content is as follows:First,by fully analyzing the complexity of the mixed-line production job shop scheduling problem,a constraint conversion scheme based on the concepts of "combined processing" and "virtual operation" is designed to transform the problem into a classic FJSP problem.Based on the actual job shop scenarios and the characteristics of agent,the structure and negotiation mechanism of MAS are designed.Secondly,in view of the shortcomings that the traditional multi-agent method only uses a single scheduling rule,which results in poor optimization results,an adaptive real-time scheduling method using contextual bandits is proposed.After learning,each agent can choose the most suitable scheduling rule according to the state of the environment to achieve better optimization of scheduling.Then,the common disturbance events in the mixed-line production job shop are analyzed.Based on the method proposed,a mechanism using machine learning is designed to solve disturbance events efficently.Finally,a prototype system is designed using JADE and My SQL.This paper also introduces its practical functions and verified the effectiveness of the system. |