Eigenfaces vs. Fisherfaces: Recognition Using Class Specific Linear Projection
IEEE Transactions on Pattern Analysis and Machine Intelligence
Nonlinear system fault diagnosis based on adaptive estimation
Automatica (Journal of IFAC)
Limitations of nonlinear PCA as performed with generic neural networks
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
This paper presents a fault detection and isolation method based on the design of a non linear PCA model and a Fisher Discriminant Analysis (FDA). A new fault detection approach based on the estimation of the prediction error (SPE: Squared Prediction Error) by the non linear PCA model is proposed. This method associates an adaptative thresholding with the study of the dynamic of the SPE. It allows to define several operating regions. The fault isolation is based on the pairwise FDA analysis applied to a class without fault and each class with fault. In this study, this new diagnosis method is validated in simulation on a quadruple-tank process. Three types of fault are simulated in a sensor: a drift, a bias and a breakdown of sensor.