Cooperative Control For Multiple CSPS System Based On Part-position Information

Due to increasing demands of the market and rapid development of technology, production system in manufacturing enterprises has experienced evolution from mass-production featured assembly line system to precise-processing featured cell production system, and then to flexible-production featured flexible cell system. As an important abstract model of automation production line system, conveyor-serviced production station (CSPS) model is widely used in many factory control areas. The research on multi-station CSPS system could provide considerable economic benefits, In this thesis, we study both the coordination control of multi-station CSPS system and its optimization under production-marketing integration model.Firstly, we concerned the cooperative control of CSPS system with infinite bank capacity. The vacancies of the bank and position information of partial parts were introduced to describe the state of the system, and optional actions of a station include unloading and servicing. Then we established the mathematical model of the system. A part-flow evaluating method is used to present a part-flow information based cooperative control mechanism for multi-station CSPS system. Then, performance potential based Q-learning algorithm was adopted to derive optimal control policy for the system. Learning effects with either fixed or Erlang-distributed processing time were analyzed through simulation experiments, and evaluation results show the effectiveness of our proposed mechanism.In addition, we further concerned the stochastic demands of customs in the decision process and studied the collaborative control of multi-station CSPS system under demand-driven mode. Both the physical model and mathematical model of the system are built, and three wake-up mechanisms were proposed to wake up the sleeping stations when custom was coming. Then, Q-learning algorithm was adopted to derive optimal control policy by simulation. The performances of the system, including custom attrition rate and load balance, were analyzed, which provided some instructions for the selection of wake-up mechanisms.