Advancements in robust fault reconstruction using sliding mode observers

With the massive growth of technologies used in the industry, more is expected from the control systems. Issues such as safety, efficiency and economical operation of the systems and the equipment are of utmost importance as these systems are expected to perform complicated and complex tasks. Hence, there is an essential need for fault detection and isolation. This thesis describes the developments and advancements in the use of sliding mode observers for the purpose of robust fault reconstruction. Existing work has shown that the effect of the disturbances onto the fault reconstruction was minimised in an L2 sense; but complete disturbance decoupling was not guaranteed. The first contribution of this thesis is the investigation of the conditions that enable actuator (input) faults to be reconstructed independently of the disturbances. The conditions are found in terms of the original system matrices. Then, the next contribution is the extension of those results to the case of sensor (output) faults. Following that, using the results of the first contribution, a scheme was developed using cascaded observers to enable disturbance decoupled fault reconstruction for a more general class of systems. Finally, this thesis presents a method to reconstruct faults in cases where the first Markov matrix is rank deficient. The efficacy of these schemes is demonstrated by simulation examples. Accompanying CD-R not available with eThesis.