ROBUST MODEL PREDICTIVE CONTROL USING CONSTRAINT RELAXATION FOR FAULT TOLERANCE

Mariana S.M. Cavalca, Roberto K.H. Galvão, and Takashi Yoneyamaa

Keywords

Robust predictive control, linear matrix inequalities, faulttolerantcontrol, physical and operational constraints

Abstract

As practical applications of model-based predictive control become disseminated throughout the industrial environment, much concern has been raised with respect to the issue of guaranteeing adequate tolerance to faults in the process. Even when robust model-based predictive control is used, eventual mismatches due to faults can lead to a significant performance degradation of the control loop or even to the non-feasibility of the optimisation problem. In order to contribute to the solution of this inconvenience, the present paper proposes a method to accommodate faults by switching between robust controllers. Although each fault-specific controller is designed to admit polytopic uncertainties, the switching from one model to another may lead to unfeasibility of the underlying constrained optimisation problem. Therefore, a technique to relax the operational constraints on the control variables is conceived to mitigate the problem of unfeasibility. A case study using numerical simulation is included to illustrate the proposed methodology.

Important Links:



Go Back