© . This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/

This paper considers detectability of deviation of sensors from their nominal behavior for a class of linear time-invariant discrete-time systems in the presence of bounded additive uncertainties. Detectable sensor faults using interval observers are analyzed considering two distinct approaches: invariant-sets and classical fault-sensitivity method. It can be inferred from this analysis that both approaches derive distinct formulations for minimum detectable fault magnitude, though qualitatively similar. The core difference lies in the method of construction of the invariant set offline in the former method and the reachable approximation of the convergence set using forward iterative techniques in the latter. This paper also contributes in giving a formulation for minimum fault magnitudes with invariant sets using an observer-based approach. Finally, an illustrative example is used to compare both approaches.

Peer Reviewed