Closed loop identification of a distillation column

To improve the closed-loop operation of a distillation column, a model is needed for control design. To obtain the data, a closed-loop experiment is designed and executed. Models are obtained by applying the direct identification method and the two-step method to the closed-loop data. Both output error models and linear regression schemes based on orthonormal basis functions (ORTFIR) are used. It appears that a combination of the two-step method and an ORTFIR model in the first step gives good results. The results are compared with the currently implemented model, and show considerable improvement in prediction capability. The new model is currently used to redesign the control loop.<
>