Actuator fault detection and isolation of nonlinear systems by robust fuzzy observers

Abstract

This paper presents a model-based technique for fault detection and isolation (FDI) of actuators of a benchmark which schematizes a hydraulic process made up of three tanks. Takagi Sugeno's model approach is used for describing the dynamic of the system. In the same way, the fuzzy membership functions used for constructing Takagi and Sugeno's model are combined with local unknown input observers to form robust fuzzy observer. Sufficient conditions for the existence of this fuzzy observer are derived. The stability as well as eigen-value constraints conditions are presented and solved in the LMI framework. For the observer gives a good estimation without amplifying noise and with a convergence faster than the dynamic of the system a eigen-value assignment is necessary. Robust residual signals, generated by these fuzzy observers robust to unknown inputs are dedicated to supervise actuators. These residuals are sensitive to faults acting on one actuator and are also insensitive to faults on the others by considering faults such unknown disturbances. This permits to carry out directly the isolation of the faulty actuator.