Robust fault detection by simultaneous observers

Abstract

This thesis addresses the problem of fault detection and isolation in linear systems based on unknown input observers. Functional disturbance decoupled observers which estimate specified or unspecified linear functions of system states regardless of the disturbances are first studied. Necessary and sufficient condition for the existence of such observers are presented. The investigation is extended to simultaneous disturbance decoupled observers where multiple systems are observed by a single disturbance decoupled observer. The application of disturbance decoupled observers to fault detection and diagnosis are explicitly outlined, and a new scheme that is based on simultaneous unknown input observers is proposed to enhance the already existing schemes. Finally, a detailed simulation example is carried out to examine the utility of the proposed scheme.