Optimal Intermittent Fault Diagnosis for Discrete-time Systems

In this paper, an intermittent fault detection and estimation method is proposed for a class of linear discrete-time stochastic systems. Firstly, a Kalman filter based residual generator following a generalized likelihood ratio based residual evaluation scheme is developed to detect the occurrence and disappearance of the intermittent fault. Secondly, an augmented optimal Kalman filter is designed to estimate both system state and intermittent fault simultaneously. The filter gain is composed of two matrices including a free part decoupled for the intermittent fault and a constrained part influenced by the intermittent fault. A Lagrangian approach is used to minimize the error covariance of augmented state and then to obtain the constrained filter gain. Finally, a numerical example is provided to verify the effectiveness of the proposed method.