A sliding mode fault detection scheme for corrupted measurement data exchange in a network of dynamical systems

This paper considers the problem of detection of a specific class of faults/failures in a network of dynamical systems, whereby an agent communicates corrupted measurement data to its neighbouring agents, even though the local measurement information at the agent itself is not faulty/corrupted. A network of linear dynamical systems is considered in which the interconnections among the dynamical systems occur via information exchange due to the use of distributed consensus type control laws. An approach based on sliding mode observers is developed to address this problem. A distributed network of sliding mode observers is proposed to detect this specific class of faults. The proposed fault isolation scheme is however centralised in nature. The complexity associated with the design of the observers is of the order of the node level dynamical systems and not related to the size of the network. The approach is demonstrated on an example taken from the existing recent literature.