Effects of Contaminated Data on the Performance of Self-starting Control Charts for Rare Events

Authors: Eralp Dogu, James Benneyan
Healthcare Systems Engineering Institute, Northeastern University, Boston MA, USA

Keywords: Process monitoring, Self-starting control charts, Rare Events, Contaminated Data

Self-starting control charts eliminate the phase I/phase II distinction and use successive process readings to update the parameter estimates and simultaneously monitor the process. Although often they are advised as simple procedures for short run processes, this practice is not always advisable, especially for phase I start-up data of unknown stability and for data experience slow drifts. Contamination could mask normal process behavior and affect the performance of the self-starting method. We assessed the performance of these control charts based on simulated data scenarios, indicating significant deterioration in detection performance.

2015 QPRC