A predictive actuator fault-tolerant control strategy under input and state constraints

The paper deals with the design of a robust predictive fault-tolerant control for linear discrete-time systems with an application of the quadratic boundedness theory and an associated robust invariant set. The main problem is to maintain the state of the system inside the robust invariant set obtained under asymmetric input and state constraints. The proposed strategy relies on a three-stage procedure, which is based on adaptive fault estimation as well as robust and predictive controller. The fault-recovery procedure is initiated with fault estimation and then the fault is compensated with a robust controller. In a case when robust fault compensation fails, i.e. the current state does not belong to the robust invariant set, a suitable predictive action is started. The main goal of this action is to generate control allocation enhancing the robust invariant set. This appealing phenomenon makes it possible to enlarge the domain of attraction of the possibly faulty system. The final part of the paper shows an illustrative example regarding a two-tank system.