Design of a model predictive-based fault estimator for faulty nonlinear switched dynamics with guaranteed recursive feasibility

This research presents a defect estimation framework employing a model predictive approach for switched nonlinear systems. The method integrates system states and failure signals into an augmented state-space model. The squared Euclidean norm of the estimation error over the prediction horizon serves as the performance metric, minimised through a set of linear matrix inequality constraints to ensure asymptotic stability. The estimator design enforces constrained estimation error to remain within a defined threshold, enhancing robustness. Recursive feasibility is analytically demonstrated at each time step. The approach is validated on a continuous stirred tank reactor, a key component in the petrochemical industry. In fault-free conditions, with estimation error maintained below 0.05. The estimator effectively detects and quantifies both constant and time-varying faults. The system's energy function consistently decreases, confirming the asymptotic stability of the estimation error dynamics and supporting its application in fault-tolerant control.