Robust Fault Detection Using Set Computation. Application to a two-tank system

This paper presents a fault detection method based on set membership techniques applied to a nonlinear two-tank system. A set-valued observer (SVO) is implemented to estimate the state. Interval analysis is used to manage uncertainties in the system. As the model is nonlinear, a simple propagation of uncertainties is not sufficient. Indeed, in nonlinear systems the solution domain is often overestimated. To get a good detection, set-computation algorithms have to be implemented. Several methods have been used in the literature to reduce the solution domain such as ellipsoids, zonotopes and subpavings. In this paper, the set will be approximated through subpavings algorithms. ImageSp and SIVIA will be used to this purpose.