Sensors FDI scheme of Linear Discrete-Time Periodic Systems using Principal Component Analysis

In this paper, Sensors FDI scheme of Linear Discrete-Time Periodic Systems using Principal Component Analysis (PCA) is proposed. The idea of this FDI scheme is to model the Linear Discrete-Time Periodic System with multiple models PCA denoted T-PCA (T is the period). For this approach, a database of measurements collected on a system in normal process operation and the period T of this system are required. The faulty sensor is detected by comparing the observed behavior given by the measurement and the expected behavior given by the T-PCA model. The variable reconstruction approach allows isolating the faulty sensors. For our numerical example, this FDI scheme allows us to detect and isolate sensors faults and then shows the effectiveness of our approach.