Data-based design of centralized PID controllers for decoupling control of multivariable processes

This paper presents a data-based design method of decoupling PID controllers for multivariable processes. The controller design directly exploits experimental process data without identifying process models. By employing the virtual reference approach, a centralized controller that achieves ideal decoupling control is first synthesized. Then, each controller element is approximated to PID structure. The control design depends on the reference model that is optimally specified according to the attainment of the reference model and the effectiveness of decoupling. Because the controller design is applicable to closed-loop data, it is preferable in industrial applications and the method can be used to improve existing underperforming controllers. Simulation studies demonstrate the effectiveness of the proposed design methodology.