VILLAR, S. B. B. L.; http://lattes.cnpq.br/2025427670049567; VILLAR, Savana Barbosa de Brito Lélis.
Resumen:
Metamodels have been used in many engineering applications, rigorous mathematical models to approximate when their computational codes require too great a time so that its practical use is possible. In this context, there was an application of the Kriging model to obtain metamodeling results of a propylene distillation separation process. This work included the use of Artificial Neural Networks as a comparison parameter between metamodels. The procedure involves the plan Latin Hypercube Sampling, selection of the type of metamodel, parameter estimation and validation. The performance of the metamodel was compared with results obtained from the rigorous model belonging to the process simulator Aspen Plus®, where the prediction of the data showed with great precision and significantly less computational effort. Another important contribution of this work is the development of methodology for optimization based on the prediction of data through Kriging metamodel using fmincon function of Matlab software and compared to optimize the optimization tool Aspen Plus®, reaching results for minimized thermal loads of reboilers the three-column distillation and obeying the purity of the product restrictions and boilup rate.