A multiple-model approach to fault tolerant tracking control for non-linear systems

This paper presents a new approach to active fault-tolerant control for non-linear systems, based on fault estimation and compensation for time-varying faults. The work is a motivation for utilising the extension of the well known so-called fast adaptive fault estimation strategy to non-linear systems via a multiple-model strategy involving Takagi-Sugeno (T-S) fuzzy observer-based estimation and linear model-reference state tracking control. Moreover, the proposed strategy handles the case of unmatched uncertainty in the tracking error dynamics. The stability proof of the appropriate non-linear augmented system is derived via a Lyapunov condition and formulated as an efficient one step Linear Matrix Inequality (LMI) problem constraining the time-varying system eigenvalues to lie within specified regions in the complex plane. The fault compensation strategy is illustrated using a non-linear example of an inverted pendulum which includes Stribeck friction and a loss-of effectiveness actuator fault.