Implementing Batch Means Control Charts for Manufacturing Big Data

Control chart using big data collected from sensors can detect small shift very effectively. However, applying the Shewart chart directly to these data leads to many false alarms, since the process big data is auto-correlated. This paper presents a method to construct batch means control charts that can be easily applied to process big data with autocorrelation. Through a simulation study, this paper presents best control chart plans according to the degree of autocorrelation in terms the number of observations spaced between batches and batch size. The applicability of the results of this study was confirmed by a practice case study of acceleration data using a ‘physics toolbox’ application on a smartphone. Opinions on further big data control chart education are also presented.