Fast actuator fault-tolerant control for a class of nonlinear sampled-data systems via deterministic learning.

In this paper, we investigate the fast fault-tolerant control (FTC) problem based on deterministic learning approach (DLA) for a class of nonlinear sampled-data systems with actuator faults, which consist of two stages: incipient faults with small magnitudes and faults with larger magnitudes. First, a learning controller and a learning identifier are constructed. Based on DLA and the exponential stability of a class of linear time-varying (LTV) discrete-time systems, the control knowledge and the diagnosis knowledge of the actuator faults are obtained. Second, a set of controllers and a set of diagnosis estimators are constructed based on the learnt control and diagnosis knowledge. When an incipient actuator fault occurs, fast fault detection and isolation (FDI) can be achieved using the diagnosis estimators. Then, the pattern-based FTC scheme is implemented to improve the control performance. When the small fault grows to a larger one, the rapid FDI and FTC are implemented again, providing fast responses to the occurred larger fault. The advantages of the proposed method are that: (i) a simple adaptive learning controller with the filtering technique is designed, in which the exponential convergence of the tracking error and parameter estimation errors can be achieved simultaneously; (ii) the sensitivity to small actuator faults is enhanced, and the fast FTC to larger actuator faults is achieved by utilizing the learnt knowledge. Simulation results are also included to illustrate the effectiveness of these schemes.