Process Analysis with Data Segments

Abstract

Abstract Control charts are applied in running a process under statistical control. Process capability analysis is conducted retrospectively on data to characterize a process under control and set up the control limits of a control chart. In this article, we expand on the process capability analysis by considering time‐related patterns. Specifically, we focus on data segments that can show up in the data collected during process capability analysis. If such segments are detected, and investigations reveal the causes for such segments, the analyst might decide to focus on a data segment representing a steady‐state stable process, as opposed to the full data collected for process capability analysis. In this context, one will aim at identifying transient periods and not include them in the computation of control limits. We provide here two methods for identifying such segments. One method is based on decision trees. The second method is applying functional data analysis.