Dynamic system failure detection and diagnosis employing sliding mode observers and fuzzy neural networks

A strategy for dynamic system failure detection and diagnosis is proposed, based on sliding mode observers, employed for residual generation with discrimination among the error subspaces, and a fuzzy neural network used for pattern classification. A control reconfiguration scheme is proposed, employing both the fault diagnosis information and the robust observer generated data. The resulting structure has been evaluated in a simulated D.C. electric drive.