Benefits of feasibility constrained sampling on unit operations surrogate model accuracy. (May 2023)

• Record Type: Journal Article
• Authors: Zinare Mamo, Tesfayesus; Di Pretoro, Alessandro; Chiari, Valentina; Montastruc, Ludovic; Negny, Stéphane
• Is Part Of: Computers & chemical engineering. Volume 173(2023)
• Publication Date: 2023-05
• Subjects: Surrogate modeling -- Design of experiment -- Flash -- Distillation -- Chemical processes; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2023.108210
• Languages: English
• ISSNs: 0098-1354

Abstract

Abstract: Due to the increasing amount of data to be treated and, thus, to the need of more computational effective strategies, surrogate modeling has become a topic of major interest during the last decades. In the modeling procedure the data set generation plays a role of critical importance in terms of accuracy of the final solution. When dealing with chemical processes, some points generated during the sampling phase could fall into a physically unfeasible region of the domain. That is why, the design of experiment step needs to be integrated with constraints able to describe the feasibility limitations of the system. By means of a dedicated interface between ProSim {\protect \relax \special {t4ht=®}} process simulator, ALAMO {\protect \relax \special {t4ht=®}} modeling software and Matlab {\protect \relax \special {t4ht=®}} for data processing, this research work assesses the benefits of constraints and subset-based approach on the accuracy of the model in terms of outlet streams and economic indicators for separation units. Highlights: Surrogate modeling is able to decrease computational effort with good result accuracy. Dividing the sampling domain into subsets reduces the error. Including feasibility constraints improves the quality of the output values.