Fault detection design of networked control systems

Abstract

In this study, the fault detection design of networked control systems (NCSs) considering time varying transmission delays, packet dropouts and quantisation errors is proposed. The design of the residual generator is formulated in the H∞ framework, where the transmission delays are described as polytopic uncertainties, quantisation errors are modelled as norm bounded uncertainties and packet dropouts are described as a binary Bernoulli process. The dynamics of residual generator is shown to be governed by a Markov jumping linear system with uncertainties, and then the residual generator is designed to be sensitive to system faults and robust against network-induced effects by applying a reference model strategy. A new residual evaluation scheme for NCSs is also proposed, where the absolute value of each residual signal is selected as the evaluation function and the threshold is computed by considering the mean value and variance of residual signals. In this way, the upper bound of the false alarm rate is ensured.