Qualitative trend analysis of the principal components: application to fault diagnosis

Abstract Quick fault detection and diagnosis is very important for effective process operations management. Qualitative trend analysis (QTA), a data-driven semiquantitative technique, has been widely used for fault diagnosis (FD). Though QTA provides quick and accurate diagnosis - the increase in computational complexity of QTA with the increase in the number of sensors used for diagnosis - may prohibit its real-time application for very large-scale plants. In most of the chemical plants, the measurements are highly redundant and this redundancy can be exploited by performing principal component analysis (PCA) on the measured data. In this paper, we present a PCA-QTA technique for fault diagnosis in large-scale plants to reduce the computation time. Essentially, QTA is applied on the principal components rather than on the sensor data. The results are quite promising for the Tennessee Eastman (TE) process