Diagnosis of Linear Systems Based on Sliding Mode Observers

Abstract

In this paper, we consider the problem of the functional diagnosis of technical systems described by linear differential equations of state with constant coefficients in the presence of disturbances. Sliding mode observers are used to solve problems of detecting, isolating, and identifying faults. The proposed approach is based on the idea of constructing a reduced (having a smaller dimension) model of the original system with the selective sensitivity to faults and disturbances, based on which a sliding mode observer is built. Depending on the specific properties of the constructed model, three options (in each of which a diagnostic problem is solved): accurate fault identification, fault detection, and approximate fault identification are considered. Apart from reducing the complexity of diagnostic tools, the proposed approach allows reducing the limitations imposed on the original system for constructing sliding mode observers.