Clustering Data Procedure for the Prediction of the Recovered Volume of the Light Gasoil of a Visbreaking Column

In this work the model identification of a visbreaking column for the estimation of the recovered volume at 360 °C of Gasoil is considered and a filtering procedure for the selection of the identification dataset is presented. A high valuable product for the visbreaking process is the light gasoil; its purity can be measured by the recovered volume at 360 °C and, for control purposes, an on-line estimation of this property is very important. In this paper a new procedure for predicting the light gasoil recovered volume is presented; the approach is based on the use of a clustering Fuzzy C-Means algorithm for the selection of the input data used in the identification process. Results are presented which prove the goodness of the proposed procedure and the reliability of the estimated model in the prediction of the gasoil recovered volume.