Minimum-Variance Fault Isolation Observers for Discrete-Time Linear Systems

In this paper, we provide two methods for the observer based fault isolation of discrete-time linear systems. Both rely on designing a single, specifically parametrized fault isolation observer (FIO). In contrast to classical bank of observers approaches, this leads to a low implementation effort. Arising degrees of freedom in systems with redundant measurements are characterized and exploited to optimize robustness of the generated residuals if the system is subject to noise. The first method is based on adding structural constraints to the classical Kalman filter, and the second approach provides a solution based on linear matrix inequalities (LMIs). A comparison is given and the results are applied to an example system.