Norm based residual evaluation and threshold computation

Abstract

In this and the next two chapters, we shall study residual evaluation and threshold computation problems. The study in the last part has clearly shown that the residual signal is generally corrupted with disturbances and uncertainties caused by parameter changes. To achieve a successful fault detection based on the available residual signal, further efforts are needed. A widely accepted way is to generate such a feature of the residual signal, by which we are able to distinguish the faults from the disturbances and uncertainties. Residual evaluation and threshold setting serve for this purpose. A decision on the possible occurrence of a fault will then be made by means of a simple comparison between the residual feature and the threshold, as shown in 9.1.