Fault Detection of Uncertain Dynamic Systems

This paper deals with fault detection methods taking model uncertainties into account. It focuses on dynamic and structured uncertain models, where parameter uncertainties are described by bounded variables. In order to de-couple residuals from unknown physical variables, a parity space approach is proposed, where the parity matrix depends on uncertain parameters. In this way, residuals are not necessary equal to zero in the fault free case, but can take a set of feasible values. Two methods are proposed and compared for constructing consistency tests. The first one needs to compute value sets of residuals; whereas the second one is based on the set-membership inversion problem.