An adaptive T2 chart for multivariate process monitoring and diagnosis

Hotelling's T-2 chart is one of the most popular multivariate control charts for monitoring multiple variables simultaneously. The identification of response variables from T-2 charts is an area that is currently receiving considerable attention. This paper proposes an adaptive T-2 chart that combines process monitoring and diagnosis in a unified manner. The proposed procedure has a close relationship with the U-2 chart, but it is data oriented and does not have a priori knowledge of the potential shifts space. It can adaptively capture shift information from the sample data to construct the U-2 test statistic and is shown to be very competitive with other alternative charts for multivariate statistical process control in terms of both mean monitoring and fault diagnosis.