Robust controller design methodology for multivariable chemical processes

Abstract In this work, a methodology for a controller design of multivariable linear time invariant systems is presented. The singular values of the return difference operator and the time constant for each control loop can be easily specified through a set of rational functions in the Laplace domain. The frequency response of an ideal controller is directly obtained, allowing the implementation of different controller structures (PI, PID etc.). First, the discussion is focused on some new results of the perturbation theory which constitute the basis of the proposed methodology. Later, a complete analysis of the resulting control scheme is presented. Its robustness with respect to stability, performance and interactions as well as the normality of the closed-loop system is demonstrated. Some implementation aspects for the formulation of the ideal controller matrix are discussed. Approximation of this ideal controller to different realizable structures is then discussed. Finally, a case study of the dual-composition control of two typical distillation columns currently found in the literature is presented.