Optimal Informative Data Selection for Historical Data Driven Process Identification

In process identification, data used for mod-eling must be informative enough over the set of the used model structures. However, when using historical data for process identification, data informativity is not always guaranteed. Thus, in this paper, an optimal method for selecting informative data from a non-informative historical dataset is developed. Given a model structure, the proposed method maximizes the number of extracted data rows while considering the extracted subset informativity as a con-straint. Accordingly, an objective function is constructed and a binary genetic algorithm is used for its optimization. In a case study, the proposed method is used to extract informative data from a database of an industrial thickener. The obtained results show the capability of the proposed approach to extract a maximum volume of informative data from a non-informative dataset.