Universal predictive Kalman filter-based fault estimator and tracker for sampled-data non-linear time-varying systems

The universal state-space adaptive observer-based fault diagnostics/estimator and the high performance tracker for the sampled-data non-linear slowly time-varying system with unanticipated decay factors in actuators/system states are proposed. The optimal linearisation technique is used to obtain the exact (local) linear model at each operating state for a non-linear system, so that the actuator and state fault detection and performance recovery of a sampled-data non-linear time-varying system can be accomplished. Additionally, an improved Kalman filter-based adaptive observer is proposed to achieve a better estimation-based performance recovery than the conventional one. A residual generation scheme and a mechanism for auto-tuning switched gain is also presented, so that the proposed methodology is applicable for the fault detection and diagnosis (FDD) for actuator and state faults to yield the high tracking performance recovery. For practical implementation, this study also takes advantage of the merit of digital redesign methodology to convert a theoretically well-designed analogue controller/observer with a high-gain property into its corresponding low-gain digital controller/observer without possibly losing the high tracking/estimation as well as FDD performance recovery. Examples are given to illustrate the effectiveness and performances of the provided methodology.