An integrated methodology for plant-friendly input signal design and control-relevant estimation of highly interactive processes

A novel methodology that integrates a directional multisine input design procedure and control-relevant parameter estimation is developed, leading to desirable models for the control of highly interactive multivariable process systems. A priori information on system directionality is utilized in the input design procedure and recognized in the subsequent parameter estimation step, which consists of control-relevant curvefitting of frequency responses obtained from identification data. The integrated methodology is demonstrated in a case study of a linear distillation column. The case study showed that a combined approach involving signal design and model estimation is superior to the conventional identification approaches for systems characterized by strong process interaction and ill-conditioning. The integrated methodology demonstrates its efficiency of estimating singular values precisely with desirable gain directionality under noise-free and noisy conditions. This is an abstract of a paper presented at the AIChE Annual Meeting and Fall Showcase (Cincinnati, OH 10/30/2005-11/4/2005).