Practical implementation of fault detection scheme in a three tank system

Abstract

This thesis presents a practical implementation of fault detection scheme in a real system by using residuals signals. These residuals have been generated applying the method of a model-based fault diagnosis. Different approaches have been studied and the corresponding algorithms developed. The object of the investigation is the three tank system in which different kind of methods of fault detection were performed. Various scenarios in which faults are simulated in actuators, sensors or components are analyzed considering also the presence of noise. Due to the inherent system characteristics it presents four possible work regions which are usually not taken into account in literature, but studied in this thesis. Also special cases, when the system goes from one region to another through critical points in which the system presents singularities, are shown. Faults will be performed around the time when the system states lie in the neighborhood of the critical points and the main task is now to overcome these singularities and achieve successfully a fault detection.