Incipient fault detection and isolation in a PWR plant using principal component analysis

A method for the detection and isolation of incipient faults in field devices in industry using Principal Component Analysis (PCA) is presented. Nominal operation data typically lie on a low-dimension surface due to relationships imposed by the physics of the process and are modeled using PCA. Abnormal deviations from the surface lead to fault detection while isolation is a consequence of these deviations being in different directions for different faults. A steam generator in a pressurized water reactor (PWR) is used for demonstration.