Fault-diagnosis by disturbance decoupled nonlinear observers

A novel concept of nonlinear unknown input observers is utilized for component and actuator fault detection and isolation in a class of disturbed and uncertain nonlinear dynamic systems. The dynamic process under consideration must be modeled in such a way that disturbances as well as model uncertainties are expressed in terms of unknown input signals. The faults to be detected are also modeled as additional input signals. If the model satisfies certain conditions, it can be transformed by a change of variables into a form that remains unaffected by the unknown inputs but still reflects occurring component or actuator faults. This model description is said to be disturbance decoupled. Subsequently, a nonlinear fault-detection observer can be designed on the basis of this model. This observer will be robust to the disturbances as well as to the model uncertainties. The output of the observer can be used for state estimation as well as for the generation of so-called residuals indicating the time and location of occurring faults