Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
DETERMINATION OF VOLUMETRIC FRACTION IN ANNULAR THREE-PHASE FLOW USING ARTIFICIAL NEURAL NETWORKS
Submission Author:
Cristhian Alvarez Pacheco , SP , Brazil
Co-Authors:
Cristhian Alvarez Pacheco, Carlos Mauricio Ruiz Diaz, MARLON MAURICIO HERNANDEZ CELY, Oscar Mauricio Hernandez Rodriguez
Presenter: Cristhian Alvarez Pacheco
doi://10.26678/ABCM.ENCIT2022.CIT22-0278
Abstract
A model based on artificial intelligence centered on an artificial neural network (ANN), capable of predicting the volumetric fraction in a three-phase flow, is developed. The study is based on artificial data from the literature, which were obtained with a non-intrusive gamma-ray technique. A simulation was developed with the mathematical code MCNP-X, based on the Montecarlo method that allows the configuration of two gamma radiation emitting sources, formed by the Europium isotope (Eu^152) and the Celsium isotope (Cs^137) in conjunction with a Nal scintillation detector (TI), which has high sensitivity to radiation. After performing the gamma-ray simulation, the total energy values emitted by Eu^152 and Cs^137 through the pipe with an annular flow pattern inside were taken, which were defined as input data in the adaptative neurofuzzy inference system (ANFIS). These data were defined as input data in the ANFIS, in order to obtain predictive values of the volumetric fraction of the phases. The structuring of the ANN was developed with Matlab software, where the inputs were those used by the authors in ANFIS, developing different configurations for the hyperparameters, in order to generate predictive values of the mentioned volumetric fractions. The statistical parameters defined to determine the best predictive behavior of the ANN models were the relative error percentage (MRE%) and the mean square error (MSE). The best models were compared with the ANFIS results.
Keywords
Artificial neural networks, Three-phase flow, volumetric fraction, gamma-ray attenuation technique, Numerical simulation
