Kanban-controlled serial manufacturing systems have been popular in Japan for many years. The analytical intractibility of such systems makes simulation and heuristics essential in studying them. This paper develops some theoretical results-reversibility and dominance-that characterize the dynamics of these systems, provides insight into their behavior and helps greatly reduce the simulation effort needed to study them. These structural results also provide the basis for developing heuristics to deal effectively with particular cases. Reversibility deals with certain permutations of the machines; dominance deals with the allocation of kanbans to cells. In addition, the paper shows that if the present goal is to maximize the throughput with a given total fixed number of cards, all of the machines should be placed in a single cell. These structural results hold in general because no assumptions are made on the processing time distribution of the machines, the number of cells, the total number of kanbans, or whether or not the machines are identical. Based on the structural results, the paper also develops a heuristic for the allocation of kanbans to a balanced line. It briefly describes an implementation of this model at a laminate manufacturing plant.
