Robust fault detection and isolation with unstructured uncertainty using eigenstructure assignment

Abstract

A numerical implementation is presented for the eigenstructure assignment approach to the design of a robust fault detection and isolation observer, By formulating the problem of multiple fault isolation in an eigenstructure, we develop a parametric characterization of all allowable eigenspaces, which provides a diagnostic observer capable of multiple fault isolation. In addition, a well-conditioned design of the allowable eigenvectors is shown to reduce the effect of the sus;em unstructured uncertainty. The resulting diagnostic observer is robust to unstructured modeling errors without losing fault detection sensitivity. A procedure for the synthesis of a reduced order diagnostic observer is also presented to increase the computational efficiency of the proposed observer.