Feedback Controller Performance Assessment: Diagonal Interactor

Abstract

Although minimum variance control is not practically desirable owing to its poor robustness and/or excessive control effort requirement, it does provide an absolute lower bound on the process variance. This lower bound naturally serves as a useful benchmark to evaluate current control loop performance if reduction of process variation is the control objective. Such a control loop performance measure provides guidelines and useful information for control engineers when they design, tune or upgrade controllers or control strategies. If the best performance cannot satisfy the requirement, alternative control strategies such as implementing feedforward control and/or reducing dead time may be necessary. For a number of industrial processes (particularly pulp/paper processes), reduction of process variation is the main objective in controller design. Performance assessment with minimum variance control as the benchmark is, therefore, particularly useful for such processes. In fact, the first application of the performance assessment technique was on a paper machine (Astrom 1967), but most industrial processes are inherently multivariate in nature so performance assessment with multivariate minimum variance control as the benchmark is more desirable. This chapter is an extension of Chapter 6 by considering closed-loop performance assessment of multivariate processes with the diagonal interactor matrix.