Fault Detection and Isolation (FDI) procedures are being increasingly used for monitoring and diagnosis in the control area. In contrast to signal-based approaches, which merely rely on an analysis of the measurements, model-based FDI-procedures utilize additionally a model of the system under investigation. By linking model and measurement algorithmically, specific parameters can be calculated which indicate type, location and extent of the fault.
In this paper, a concise introduction to model-based FDI-procedures is given. A specific algorithm is explained more thoroughly: the Extended Kalman Filter (EKF), which allows for simultaneous estimation of states and parameters of linear and nonlinear systems. A modal formulation of the EKF (MOKF) is presented, which is based on a modal reduction of the system equations. Furthermore, it is explained how EKF and the Instrumental Variables method can be combined.
The model-based procedures will be applied to a simulated rotor, and to a rotor test rig modelled by finite elements in order to diagnose a crack in the shaft.