Quality-related Fault Detection Approach Based on Dynamic Total Principal Component Regression Component Regression

On the basis of the structure of auto-regressive moving average exogenous (ARMAX), we propose a dynamic total principal component regression (DT-PCR) method for dynamic performance of quality-related fault detection. We form the input augmented matrix in the method based on the delay value of the input. The augmented matrix is divided into two orthogonal parts, namely, quality-related and quality-unrelated. We design a simple fault detection strategy based on statistics in two subspaces that correspond to the two parts. The output prediction accuracy of DT-PCR is better than that of former methods. The prediction and fault detection performance of the proposed approach are proved by a numerical example and the Tennessee Eastman process through a comparison by using total partial least squares (TPLS).