A distributed production control for intelligent manufacturing systems
Global competition and advances in technologies have resulted in some fundamental changes in the manufacturing environment. The fast rate of change is increasing without any pause in sight. A practical control/scheduling strategy which can handle complexity, cope with the changing environment, and accommodate changing multi-objectives with the aid of advanced production technologies and facilities is worth exploring.^ In this thesis, we present a production control framework that utilizes distributed decision making and distributed information flow and is based on a price and objective mechanism. The framework supports heterogeneous job objectives, admits job priorities, recognizes multiple resource types, and allows multiple step negotiation between parts and resources. Under the framework, each job enters the system with some currency and tries to fulfill its processing requirements to meet its objectives by actively bargaining with resources to purchase processing services. On the other hand, resources determine their service charge based on demands and try to sell their services to maximize their profit. In this way, the complicated coherent manufacturing control problem is decomposed into a collection of independent agents' decision-making problems. The global states and entity interactions are reflected and controlled by a price system.^ In order to ensure harmonious and effective operation, price adjustment among different resources and parts needs to be handled with care. This issue is addressed through the presentation of different distributed control schemes: a part-initiated negotiation scheme, a resource-initiated negotiation scheme, a multiple-reservation negotiation for both part-initiated and resource-initiated schemes, and a bottleneck centered, look-ahead negotiation scheme.^ A Flexible Routing Adaptive Simulation System has been built to demonstrate the flexibility and effectiveness of the proposed framework. The performance of different control strategies is analyzed and compared. The results show that the proposed frame-work provides a foundation for highly adaptive production control, and the price system provides a mechanism for system state monitoring in a distributed environment. ^
Major Professor: James Solberg, Purdue University.
Engineering, Industrial|Artificial Intelligence