Approach to Fault Diagnosis in Linear Systems by Non-Parametric Method

Abstract

The problem of fault diagnosis in technical systems described by linear dynamic models is studied. The suggested method of diagnosis is based on sliding mode observers estimating the disturbances that then is used for diagnostic observer operation. As a result, a degree of robustness of the diagnostic process is increased significantly. To realize the suggested method, three steps have to be fulfilled. Firstly, the reduced model of the system under consideration invariant with respect to the faults and sensitive to the disturbances is constructed. Based on this model, the sliding mode observer estimating the disturbances is obtained. The feature of such an observer is that a sliding motion takes place forcing the estimation error equal to zero in finite time therefore the disturbances can be estimated. Finally, the diagnostic observer sensitive to the faults and the disturbances and using the estimated disturbances as the known input is designed. This observer generates a residual which is used for decision making about faults. Since the disturbances are used now as the known input, a degree of robustness is increased significantly. Theoretical results are illustrated by practical example of the general electric servoactuator of manipulation robots under absence of the external loading moment.