Robust recursive principal component analysis modeling for adaptive monitoring

This paper proposes a robust recursive principal component analysis (PCA) modeling procedure that aims to improve the monitoring performance by detecting and identifying process changes, removing disturbances, and updating the model to reflect the operating mode change. The proposed approach was applied to an industrial fired heater. Compared with previous approaches based on conventional PCA or recursive PCA, this new procedure demonstrated improved monitoring performance. The case study shows that both the number of false alarms and the number of model updates were significantly reduced in comparison with previous methods.