Abstract
In this paper we propose a solution to the model-based fault diagnosis problem for the class of non-linear dynamic systems subjected to be described by a Takagi-Sugeno fuzzy model. A fuzzy observer is designed to estimate the system's state vector and to derive a diagnostic signal-residual. The residual is generated by the comparison of the measured and the estimated outputs. The proposed scheme has been satisfactorily tested in simulation and in a real-time benchmark given by a Two-Tank Hydraulic System.