On computing achievable fault signatures

Abstract

The knowledge of achievable fault signatures is a valuable information in designing residual generators providing structured residual sets for fault detection and isolation. We propose an ecient computational approach to determine the achievable fault signatures for a given additive fault model. The proposed procedure relies on recently developed numerically reliable nullspace updating techniques involving orthogonal reductions to Kronecker-like forms. The new procedure is general, being applicable to both proper as well as non-proper systems, and is signicantly more ecient than an exhaustive search based approach.