A response surface methodology for parameter setting in a dynamic Conwip production control system

Pull production control systems are usually implemented by means of Kanban cards. Kanban cards can be established either statically (card setting) or dynamically (card controlling). As card controlling seems to have a number of advantages over card setting, several card controlling mechanisms (CCM) have been proposed in the literature. The CCM automatically adjust the number of cards required depending on the status of the system and on the values of certain constants or parameters. Therefore, the performance of a CCM is greatly affected by the correct setting of its parameters. In this paper, we address the parameter setting in a specific CCM designed for Conwip systems. We focus on this mechanism as it has proved to outperform other CCM for a variety of scenarios. We suggest a methodology based on response surface methodology (RSM) statistical procedure for parameter setting. In order to test its suitability, we apply this methodology to a five-station line. The results show that the performance of the CCM for the predicted values of parameters is very close to the optimal solution, obtaining savings of 82% in the number of simulations, as compared to the results obtained by an exhaustive search of all possible parameter combinations.