Fault diagnosis based on set membership identification using output-error models
Résumé
SUMMARY The objective of this article is the design of a fault diagnosis method based on set membership identification for systems subject to abrupt parametric faults. The proposed method assumes an output-error, linearly parametrizable model, with unknown but bounded noise corrupting the measurement data. The set membership identification's objective is to compute (a) the smallest volume-wise ellipsoid that contains the nominal parameter vector, which concurrently resides within a data-hypersector and (b) the monotonically nonin-creasing parametric set arisen from the orthotopes that tightly outer bound the ellipsoids. The fault detection mechanism is based on consistency tests between the estimated ellipsoids and the data-hypersectors and the intersection of support orthotopes. At the sample instant of the fault detection, set-theoretic operations are applied for the subsequent fault isolation and identification. The fault isolation is based on the projections of the intersections of orthotopes, while the distance of their centers is used for fault identification. Simulations studies are used to verify the efficiency of the suggested method applied on an electrostatic microactuator subject to several abrupt failure modes.