Model-free residual generation for sensor fault detection and isolation in bilinear systems
Résumé
On-line Fault Detection and Isolation (FDI) approaches are based on fault indicator signals namely residuals. These residuals allow to verify the consistency of the known signals (control signals and measured outputs) with respect to a reference model. Classical model-based residual generation methods, namely parity space based and observer based methods, need the perfect knowledge of the model equations i.e. the structure SM and the parameters PM. However, in many practical situations, model equations are not exactly known, and very often only the structure SM is available. This paper presents a novel approach that allows to generate residuals for system of bilinear structure SM and under a stability hypothesis, but without an explicit knowledge of the values of parameters PM.