Multivariable System Identification for Integral Controllability – Computational Issues

A process model satisfies the integral controllability (IC) condition if the model can be used in a model-based controller that can be arbitrarily detuned without jeopardizing closed-loop stability. For decoupling multivariable control this requirement is equivalent to the inequality 1 ˆ Re 0 GG for the actual and estimated process steady-state gain matrices G and Ĝ . This necessitates experiments for identification of Ĝ that satisfies the IC inequality. In this work we explore, via computer simulations, computational issues related to the design of such experiments for an FCC process. The proposed approach is based on a general mathematical optimization framework we presented in prior work.