Spatial detection of manufacturing shift in mean

The development of Statistical Process Control mostly focuses on temporal data. Those techniques are not sufficient for analyzing the manufacturing information embedded with spatial structure. The monitoring process that treats manufacturing records as purely multivariate data would miss important information of spatial relationship among the records. Thus, this paper proposes a spatial-EWMA test procedure to identify process mean shift in a plane. The goal is to analyze whether there is an outbreak cluster in the investigated area. The proposed method uses relative distances among observations to construct test statistics. The observations which are far away from the outbreak center receive smaller weights since they tend to be less influential. The simulation results show that the detection power of the spatial-EWMA test procedure is preferable to the classical multivariate control chart like T² chart even in the high dimensional scenarios.