Detecting intermittent sensor faults for linear stochastic systems subject to unknown disturbance

Intermittency of sensor faults leads to wrong control commands, fault propagation and overmuch maintenance, which complicates the detection task for characterizations of unpredictability, discontinuity, non-determinacy and repeatability. In this paper, we present a novel robust detection method for intermittent sensor faults in linear stochastic systems corrupted by unknown disturbance. In order to detect all the appearing time and the disappearing time of an intermittent sensor fault, a scalar robust residual is designed by utilizing the parity space approach such that the residual gets decoupled from unknown disturbance but most sensitive to the intermittent senor fault. Based on the analysis of characteristics of the scalar residual, two hypothesis tests are proposed to detect all the appearing time and the disappearing time, respectively. Within this framework, the detectability of an intermittent sensor fault is further defined and a sufficient condition is obtained to ensure that (i) all the appearing time and the disappearing time are detected; (ii) each appearing (disappearing) time of the intermittent sensor fault is determined before the subsequent disappearing (appearing) time in a probabilistic sense. In addition, capacities of detection speed, false alarm rates and missed detection rates are theoretically analyzed. Finally, efficiency of the proposed scheme is verified through a simulation on a simplified vertical dynamic model of an F-16 aircraft.