Fault Detection Using Consensus-Based Linear Distributed Kalman Filtering

The paper addresses the discrete-time distributed Kalman filtering over a sensor network to generate system fault residuals. The problem of interest are distributed sensor data consensus filtering for systems whose dynamics, as well as the control objectives, are not decoupled. Limited to linear systems and fixed network connectivity, the state observability and fault detectability is analyzed to guarantee stability of the distributed filters and responsibility of fault residuals. The performance of fault residuals generated within distributed Kalman filtering algorithms are demonstrated in the numerical example.