Robust failure detection with parity check on filtered measurements

This paper reports a new development in failure detection based on the parity check method. The novelty lies in that the parity check is no longer performed directly on a sequence of measurements, but on a sequence of filtered measurements instead. With a modest increase in complexity, the ability to reduce model uncertainties now becomes an inherent merit of the new detection scheme. The filter gain can be determined to optimize the robustness of a set of redundancy relations with respect to uncertainties in signals/models. An example is given to demonstrate the superiority in robustness offered by the new design method