A data-driven approach for the design of feedback controllers

In this paper a simple solution to obtain a feedback controller from input - output process data is proposed. The method uses the virtual reference structure and the virtual error in order to identify a dynamic compensator (DC) that achieves the derivative behaviour of the controller. The DC, defined as an adaptive filter, achieves a precise and extended compensation of the process dynamics. The controller gain is computed in two ways: either by imposing the gain stability margin, as in the case of processes with significant dead time or with large pole excess, or by imposing the phase stability margin, for other types of processes. Both ways are illustrated in the paper by conclusive case studies.