Search

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. In this paper we develop some theoretical results—reversibility and dominance—that characterize the dynamics of these systems, provide insight into their behavior and help 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, we show that if our 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, we also develop a heuristic for the allocation of kanbans to a balanced line. We briefly describe an implementation of this model at a laminate manufacturing plant.

We describe and analyze a discipline called kanban for the coordination of cells in a large-scale manufacturing facility. There are many cells in a tandem configuration and a fixed number of cards (or kanbans in Japanese usage) in each cell. The cards circulate within the cell and their presence at designated positions signals to the neighboring cells the status of the cell's inventory; rules specify the timing of the movement of parts from cell to cell. We construct a stochastic model of the whole manufacturing facility and give our results in three parts. First, the kanban discipline is compared to the classical production discipline and it is shown to dominate the classical discipline in terms of throughput. Secondly, we give a scheme for analyzing approximately the performance of the kanban system by examining first a single cell in isolation and then combining the isolated cells through fixed-point equations. Thirdly, we report on extensive experiments and on their numerical results which are obtained from simulations and our method of analysis. The throughput-inventory relationship of the kanban scheme is observed to be superior to that of the classical discipline.

In this paper we present a mathematical programming model for the Kanban system in a deterministic multi-stage capacitated assembly-tree-structure production setting. We discuss solution procedures to the problem and address three special cases of practical interest.

There has been a great deal of interest recently in the Japanese approach to manufacturing, growing out of a concern for finding ways to reduce inventories and increase productivity. At this project's inception, its objective was to assess whether the kanban system could perform well in the manufacturing environments found in this country. Based on observations from managers visiting Japan, the project was enlarged to also assess which factors in a production environment have the biggest impact on performance—regardless of the system in use. Guided by a panel of production and inventory managers from diverse plant environments, a comprehensive list of factors thought most important to manufacturing effectiveness was constructed. The panel established low and high values for each one. These settings were considered representative of the range experienced in U.S. plant environments. The factor settings allowed a variety of representative plants to be tested with a large scale simulator. Results show that kanban, when implemented in certain environmental settings, does indeed perform exceptionally well. However, so do the more traditional systems used in the United States. Conversely, there are other plant environments in which all systems perform much worse. This suggests that the factors themselves are the keys to major improvement. Simultaneously reducing setup times and lot sizes is found to be the single most effective way to cut inventory levels and improve customer service. Shop factors of particular importance are yield rates and worker flexibility. Degree of product standardization and the product structure are also high impact factors. Less crucial than earlier believed, at least over the factor settings simulated, are inventory record inaccuracy, equipment failures, and vendor reliability. Such results suggest that the selection of a production/inventory system can be of less importance than the improvement of the manufacturing environment itself.