Fault diagnosis of Networked Control Systems: Robustness enhancement

The fault detection problem is studied in this paper for a class of linear Networked Control Systems (NCS) with communication delays. The aim is to generate residual signals which are supposed to be identically zero in the fault free case. In practice, this condition is never satisfied due to various factors such as measurements noises, model uncertainties, and induced delays in particular for NCS. The effect of unknown networked induced delays on conventional observed based residual generator is studied. It is shown that the detection performances may be deteriorated due to the sensitivity of the residuals to the delays. Two approaches for robustness enhancement are proposed: one is based on robust residual generation making use of an adaptive Kalman filter, the other relies on adaptive residual evaluation.