Robust fault detection filter design for linear uncertain systems with unknown inputs

Abstract

In this paper, a robust fault detection filter design method for uncertain systems in linear fractional transformation (LFT) formulation with unknown inputs is proposed. The basic idea is to convert the complicated ℋ-/ℋ ∞ problem to an easier ℋ ∞ model following problem. Moreover, two major improvements have been made in this research. First, the uncertain systems in LFT formulation are studied. This class of uncertain models is capable of approximating complex nonlinear dynamics. Second, a more general form of filter is employed to achieve a better fault detection and disturbance rejection performance. It involves the widely used observer-based filter as a special case. With structured uncertainties, it has been shown the robust fault detection filter design can be solved by a convex optimization condition in terms of linear matrix inequalities (LMIs). An illustrative design example is used to demonstrate the effectiveness and better performance of the proposed approach.