A novel fault diagnosis approach for a class of discrete descriptor systems

Abstract

A novel fault diagnosis approach for discrete linear time-invariant descriptor systems is proposed in this paper. There are two highlights. (1) Through model transformation, a dynamic or static sub-system described by a regular difference equation or algebraic equation is extracted from the original system, and the resulting sub-systems are coupled with faults. (2) Based on the dynamic or static relationship between the faults and the known variables, the faults are identified by the least squares algorithm. The identification error of the faults, the selection of the threshold and the problems of model uncertainties and unknown disturbances are also analysed in this paper. A simulation shows the effectiveness of the proposed fault diagnosis approach.