Reconfiguration blocks for fault-tolerant control of nonlinear systems

Abstract

Industrial processes and technological systems are becoming more and more autonomous and complex. Consequently, the demand for the safety and reliability of these systems is increasing. In this context, process monitoring and fault-tolerant control (FTC) have received a lot of attention during the last decades to provide analytical redundancy to these processes and improve their reliability. This thesis addresses the problem of FTC for nonlinear systems based on the fault hiding approach. Fault hiding consists in inserting a reconfiguration block (RB) between the faulty plant and the controller. The RB mitigates the fault effects by dispensing with the controller redesign, through recovering sensor measurements and reallocation of the control effort required by a controller that does not receive the information about the fault occurrence. Although effective, most of the fault hiding approaches available in the literature do not cover all classes of nonlinear systems, and they are sensitive to fault estimation inaccuracy because the canonical RB structures, known as virtual sensors and actuators, rely on the internal model principle. In this sense, this thesis addresses the problem of fault hiding for nonlinear systems based on novel RB structures whose parameters do not exhibit explicit dependence on the fault model. This thesis presents a novel constructive design with sufficient conditions based on linear matrix inequalities (LMIs) for guaranteeing stability recovery by fault hiding. For obtaining those conditions, three novel classes of approaches are proposed, namely: a Lyapunov-based approach wherein a Lyapunov function is obtained in a stability analysis step for the nominal system, then it is used to design the RBs for stability recovery of the reconfigured system in a synthesis step; a dissipativity based approach wherein dissipation inequalities are obtained in an analysis step for the nominal system, then they are used to design the RBs for dissipativity recovery of the reconfigured system in a synthesis step; finally, a passivation-based fault hiding approach is proposed to compute RBs with dissipativity properties which compensate for the lack of passivity of the closed-loop system due to the fault occurrence.