Kanban information

ARTICLE IN PRESS

Int. J. Production Economics 99 (2006) 102–116 www.elsevier.com/locate/ijpe

Dynamic card controlling in a Conwip system$
´ Jose M. FraminanÃ, Pedro L. Gonzalez, Rafael Ruiz-Usano
Industrial Management School of Engineering, University of Seville, Avenida de los Descubrimientos, E41092 Seville, Spain Available online 29 January 2005

Abstract The main parameter affectingthe performance of pull systems (such as Kanban or Conwip) is the number of cards employed to control the release of the material. Therefore, establishing the correct number of cards in pull systems is an important issue that can be addressed either statically (i.e. card setting), or dynamically (i.e. card controlling). In this paper we focus on card controlling in Conwip systems. First, wereview the different contributions regarding card controlling in pull systems. Then a new procedure is proposed and tested under different environments. The procedure shows to be competitive with respect to the existing ones. Finally, we compare card setting against employing the suggested procedure for card controlling. r 2005 Elsevier B.V. All rights reserved.
Keywords: Production control systems;Pull systems; Conwip; Card controlling; Card setting

1. Introduction Pull production control systems, such as Kanban or Conwip, are considered to be superior to push systems in those manufacturing scenarios where both can be applied (see e.g. Spearman et al., 1990; Roderick et al., 1992). Pull systems control work in process (WIP) and measure throughput rate or service levels. The Kanban system(see e.g. Monden, 1983) limits the WIP on each station by
This work has been supported by CICYT project DPI20013110. ÃCorresponding author. Tel.: + 34 95 448 7214; fax: +34 95 448 7329. E-mail address: jose@esi.us.es (J.M. Framinan).
$

employing a number of cards (Kanbans) for each station. A job can be processed in one station only if there are available cards in this station to be attachedto it. When this job has been processed, the card is removed and it can be attached to a new job. In contrast, Conwip systems (Spearman et al., 1990) do not limit the WIP on each station, but the total WIP in the system. To do so, a card is attached to a job when it enters the system and it is withdrawn when it finishes its processing. It is clear that establishing the number of cards greatlyaffects the performance of the system. In the case of Kanban, one must establish one card count per station. Therefore, n parameters have to be established for an n-station Kanban system. In the case of Conwip, the number of parameters is

0925-5273/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpe.2004.12.010

ARTICLE IN PRESS
J.M. Framinan et al. / Int. J.Production Economics 99 (2006) 102–116 103

one, i.e. the number of cards to be employed in the system. Establishing the number of cards in pull systems can be addressed from two different points of view (Framinan et al., 2003): (a) Card setting, i.e. given certain manufacturing conditions, employ a procedure to set the number of cards that makes the performance of the system to be acceptableaccording to some pre-defined operating measures. The number of cards obtained by this procedure is assumed to be fixed during the decision interval. (b) Card controlling, i.e. to devise rules for changing or maintaining the current number of cards depending on certain events (such as changes in the demand rate or WIP excess) taking place in the manufacturing scenario. These rules are aimed to obtain atarget performance of the system. With respect to card setting, there are several contributions regarding this topic. With respect to card setting in Kanban systems see e.g. Monden (1983), while for a review and classification of card setting methods for Conwip systems, see Framinan et al. (2003). In contrast, there are few works dealing with procedures for card controlling, including the papers by...