Observer-based strategies for actuator fault detection, isolation and estimation for certain class of uncertain nonlinear systems

Abstract

Two observer-based actuator fault isolation schemes for a class of uncertain nonlinear systems have been presented. To deal with a broader class of uncertain nonlinearities than previously considered, novel diagnostic observers are proposed, which combine two nonlinear observer design strategies, namely Thau's observer with sliding-mode observer concepts. The proposed observers are primarily designed for actuator fault diagnostic purposes. The nonlinearities that can be attacked may include both Lipschitz nonlinearities and those uncertain nonlinearities that are not Lipschitz but satisfy certain matching conditions. Design of the proposed observer boils down to solving a set of linear matrix inequalities (LMIs), which can easily be accomplished using the Matlab's LMI toolbox. Using the proposed observers, two actuator fault isolation schemes are designed. Unlike the existing techniques, using m observers (where m is the number of actuators) in the first approach, and only one observer in the second approach, the proposed schemes can isolate any number of actuator faults occurring at the same time. In addition, both the proposed schemes are capable of estimating the shape of the faults which is useful for fault accommodation purposes. A numerical example is provided to show the effectiveness of the proposed model-based actuator fault isolation strategies. The simulation results confirm that the two proposed techniques are effective in dealing with robust actuator fault detection, isolation and estimation in the studied class of nonlinear systems.