Eliminating false alarms caused by fault propagation in signal validation by sub-grouping

Many signal validation methods detect sensor faults by comparing signal measurements with their estimates reconstructed from other correlated signals. For these methods, one faulty signal may cause all the other signals that use the faulty signal as an input in signal estimation to be un-correctly reconstructed and subsequently falsely identified as faulty too. This phenomenon is called fault propagation and can produce excessive false alarms in signal validation. This paper presents a general sub-grouping technique that uses specially designed intersections between sub-groups to eliminate the false alarms caused by fault propagation. Two methods are developed based upon this technique. Demonstration shows that both methods can dramatically reduce the occurrence of fault-propagation-caused false alarms in signal validation.