Purpose – The search of the optimal economic design of the Bayesian adaptive control charts for finite production runs can be a long and tedious procedure due to the intrinsic structure of the optimization problem, which requires a dynamic programming approach to select the best decision at each sampling epoch during the production horizon of the process. This paper aims to propose a new efficient procedure implementing a genetic algorithm neighbourhood search scheme embedded within the dynamic programming procedure with the aim of reducing the computational burden and achieving significant cost savings in the chart implementation. Design/methodology/approach – The efficiency of the developed procedure has been verified through a comparison with another existing exhaustive approach working exclusively on one-sided X Bayesian control charts; then, it has been extended to the design of two-sided Bayesian control charts. Findings – The proposed procedure implementing the genetic algorithm neighbourhood search is very fast and efficient in detecting optimal solutions: it allows significant quality control cost savings to be achieved during the Bayesian charts implementation thanks to the possibility of investigating larger spaces of decisions than the existing optimization procedures. Practical implications – With reference to discrete part manufacturing, where the assumption of finite production runs is often realistic, the design and implementation of adaptive Bayesian control charts by means of the proposed procedure allows significant cost savings to be achieved with respect to the fixed parameters Shewhart charts. Originality/value – The exhaustive optimization procedure cannot be executed in a reasonable computational time when the space of decisions to select Bayesian chart design parameters significantly enlarges, which is the case of two-sided control charts. The paper documents the proposed procedure which overcomes this problem and allows the two-sided Bayesian chart to be designed and proposed as an efficient means to monitor short production runs.
|Titolo:||An efficient genetic-dynamic programming procedure to design Bayesian control charts|
|Data di pubblicazione:||2009|
|Appare nelle tipologie:||1.1 Articolo in rivista|