Identifying the Process Shift with Robust Control Charts in the Presence of Contamination

Abstract

Conventional control charts have traditionally been reliable tools for monitoring processes under the assumption of normally distributed data. However, real-world data often deviate from this idealized normality, leading to reduced charting performance and potentially causing process anomalies to go unnoticed. In this study, by integrating robust statistical estimators and innovative charting techniques, robust control charts demonstrate their capability to effectively detect process shifts and abnormalities in a variety of challenging settings. Through Monte Carlo simulation studies and a real dataset application, this research provides insights into the benefits and limitations of robust control charts. Our findings indicate that the proposed robust control charts show a notable performance in detecting data anomalies, specifically for the shift in mean, outperforming conventional charts in this regard. Comparison among the three robust location estimators via simulations, namely Huber (H) and Biweight (B) estimators as well as the proposed Biweight estimator integrating the M-Scale (BM) estimator also demonstrate its superiority in handling shifting in mean process situations.