Fault detection and isolation of sensors in dynamic systems with modelling uncertainties

This paper deals with the problem of modelling the influence of unknown time-varing parameters on linear dynamic system response. These parameters are introduced through the appropriate parts of the system matrix description in the state-space domain. The author shows that their influence can be efficiently estimated by dynamic programming, thus resulting in the so-called permissible interval of the system output. This interval defines the worst case of the uncertain parameters influence and defines the upper and lower bounds of the system response. The approach is further applied to a full-order classical observer-based scheme for fault detection and isolation of sensors, greatly enhancing its robustness. Suitable choice of the observer gain matrix is shown to significantly improve fault-detection and isolation capability. The theoretical results are illustrated by simulation of some faults modes on an aircraft example. These results verify that the proposed fault detection observer scheme can be efficiently used in dynamic systems with modelling uncertainties (errors).