Classificação de padrões de escoamento bifásico líquido-gás baseada em séries temporais de fração de vazio e técnicas de aprendizagem de máquina.

Abstract

The classification of the flow pattern is a fundamental step in many processes that involve multiphase flow, among them, estimation of mass and/or volumetric flow, calculation of void fraction, bubble size, slip factor and so on. In this work some approaches are proposed to classify liquid-gas vertical flow patterns that combine signal processing techniques and machine learning. Wire-Mesh sensor data were used to obtain the time series of void fraction. They are analyzed through statistical and time-frequency approaches, generating parameters that are later used as inputs to the algorithms based on the Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) to finally identify the pattern. The tests performed indicate that some of the proposed algorithms perform better than 90% of the correct identification. The data used are from vertical water-air flow in 51.2mm tubes.