A Novel Principal Component Analysis Algorithm for Fault Detection

In modern industry, fault detection is very important in high-cost and safety-critical processes. Principal component analysis (PCA) is an effective data-driven approach which is widely used in this field. Conventional PCA is time-invariant, which often lead to false alarms. In order to overcome the false alarms, recursive PCA algorithms have been proposed for adaptive process. However, the amount of calculation of the algorithm is very great because of frequently updated information. In this paper, a novel recursive PCA method is proposed for solving the question by introducing the standard residual method. The standard residual method serves as an indicator of the level of consistency among data, which is used to decide whether or not to update the PCA model. The proposed method is applied to fault detection and experiment results demonstrate the effectiveness of the method.