Eventos Anais de eventos
ENCIT 2018
Brazilian Congress of Thermal Sciences and Engineering
ESTIMATION OF GAS-LIQUID FLOW PATTERNS UTILIZING MACHINE LEARNING METHODS
Submission Author:
Davi Lotfi Lavor Navarro da Rocha , SP
Co-Authors:
Davi Lotfi Lavor Navarro da Rocha, André Mendes Quintino, Oscar Mauricio Hernandez Rodriguez
Presenter: Davi Lotfi Lavor Navarro da Rocha
doi://10.26678/ABCM.ENCIT2018.CIT18-0499
Abstract
In recent years, several published works use models based on artificial neural networks (ANN) to estimate the gas-liquid flow patterns; however, only few compare the accuracy of different algorithms such as support vector machines (SVM), decision trees (DT), random forests (RF) and nearest neighbors (KNN) with experimental data. Based on a large experimental database available in the literature for gas-liquid flow in pipes, consisting of different diameters, fluid properties and pipe inclination, an accuracy comparison is performed with five different algorithms. Results show that the SVM algorithm presented the best accuracy, achieving 90.03% of accuracy in correctly predicting the flow patterns.
Keywords
machine learning, Two-phase Flow, Gas-Liquid Flow, Flow patterns
