A multivariate monitoring method based on PCA and Dual Control Chart

Recently, Dual Control Chart (DCC) has been proposed for mining the changing information of a process inner variable. However, the DCC method does not consider the correlations among the variables. Further more, the G statistic of the DCC method has a cancellation problem. Though the G a statistic is proposed to overcome the cancellation problem, the changing information of G statistic is lost. In this paper, an improved Dual Control Chart based on PCA and Two-track Statistic for overcoming the two shortcomings of the DCC method. The contributions are as follows: 1) the PCA is introduced to decouple the correlations among the original variables. 2) the Two-track Statistic is explored to overcome the cancellation problem; The proposed method is applied to fault detection of Tennessee Eastman process (TE process). The monitoring results show the effectiveness of the proposed method, compared with DCC method.