Improved design of quantile-based control charts

Abstract

In a manufacturing process, deterioration in quality may be represented by changes in process variables’ distributions rather than shifts in their means or variances. In this case, when means or variances cannot reflect the real process conditions, we need to monitor whether a sample conforms to an in-control distribution. Quantile-based control charts are suitable tools which can do the monitoring. However, there are no specific procedures for the users to follow to construct such a chart. In this article, we propose a new criterion for the construction of the quantile-based control chart. With this criterion, appropriate quantiles are selected to maximize the test power of the charted statistic. We study most commonly used distributions and specify guidelines for the construction of the quantile-based control charts. Simulation results demonstrate the high efficiency of the proposed method.