AbstractsEngineering

Kanban means card or token. A kanban-controlled production system is one where the flow of material is controlled by the presence or absence of a kanban, and where kanbans travel in the system according to certain rules. The study of kanban-controlled production systems can be traced back to the Toyota Production System in the 1950s. The classic kanban-controlled system was designed to realize Just-In-Time (JIT) production. Kanban-controlled production systems, though pervasively used in industry and studied for decades, are not well understood quantitatively yet. The essence of kanban-controlled production systems is to use single or multiple closed loops to provide information flow feedback using kanbans. By doing this, the systems keep tight controls over inventory levels, while providing satisfactory production rates. The goal of this research is to study the behavior of the class of manufacturing systems with multiple closed loop structures and explore the applications in design and operational control of production systems using multiple-kanban loops. To do so, stochastic mathematical models and effcient analytical methods for evaluating the performance of systems with complex structures are required. In this thesis, we present an assembly/disassembly network model which integrates the control information flows with material flows. Blocking and starvation properties due to machine failures in a system are analyzed by establishing an effcient underlying graph model of the system. Based on the mathematical model and blocking and starvation properties, effcient and accurate algorithms are developed for evaluating the performance of systems with arbitrary multiple-loop structures. We study the behavior of multiple-loop structures and develop intuition for optimal design and operational control using multiple-kanban loops. Some practical guidelines for the design and control of production systems using multiple-kanban loops are provided at the end.