Practical application of model identification based on ARX models with transfer functions

A novel model identification methodology for ARX models based on transfer functions has been proposed. The identification approach converts transfer functions to ARX models with no approximation, except zero-order hold. Model parameters of the transfer functions are estimated directly. Model identification for process controls, especially MPCs, is of great importance for achieving the highest performance from them. However, step testing for model identification is a time-consuming task. Model identification techniques are necessary to save time for step tests. Therefore, a closed-loop identification method of multivariable systems is useful and helpful for time-saving. Herein, the proposed method, with control by model predictive controllers, is suited for a closed-loop identification technique and is applied in an industrial chemical plant.