Active Fault-Tolerant Control Strategy Based on Moving Horizon Estimation and Model Predictive Control

Abstract

This work addresses the design and evaluation of an integrated Active Fault-Tolerant Control (AFTC) framework that combines Model Predictive Control (MPC) and Moving Horizon Estimation (MHE) to achieve real-time fault accommodation in constrained multivariable systems. The proposed methodology is validated on a linearized three-tank system model, both under nominal conditions and in the presence of actuator and plant faults. The MPC formulation incorporates polyhedral state and input constraints, an ellipsoidal terminal set, and its inner polyhedral approximation to ensure recursive feasibility and stability. A parameter estimator based on MHE augments the system model to simultaneously estimate states and fault residuals online, enabling the dynamic reconfiguration of the closed-loop system. For the nominal (fault-free) scenario, the simulation results demonstrate that the proposed AFTC framework achieves accurate setpoint tracking with negligible steady-state error, while respecting all constraints. In situations of abrupt actuator fault, the method restores performance close to that obtained with perfect fault knowledge, presenting only minor degradation due to estimation delay. In cases of severe plant faults involving significant nonlinearities, such as pipe blockages and tank leakages, the proposed approach provides only partial recovery, with superior performance to the nominal-model configuration, although limited by the convergence capacity of the estimator. The study concludes that integrating MPC and MHE into a unified AFTC framework enhances system resilience to faults while ensuring constraint satisfaction. The proposed approach is particularly well-suited for linear applications and provides a foundation for future extensions, including robust and nonlinear estimations, advanced predictive control strategies, and real-time hardware implementation.