Void Signal Analysis and Gas-Liquid Two-Phase Flow Regime Determination by a Statistical Pattern Recognition Method : Fluids Engineering

Abstract

A statistical pattern recognition method was applied to the analysis of the signals of a cross-sectional mean void fraction for discriminating gas-liquid two-phase flow regimes. The analysis and discrimination were carried out based on six key flow patterns: bubble, cap-bubble, plug, froth (F_I and F_II), and annular flow. For each flow condition, 100 void signals with a recording dimension of 1 second were used and transferred to discrete data, the sampling frequency of which was selected as 100 Hz by comparison with correct recognition rates obtained from different frequencies. The magnitude of the time-averaged void fraction was partly employed supplementary to the pattern recognition method. The boundaries between the six flow regimes were determined corresponding to a correct recognition rate of 80% and were drawn on a superficial gas-liquid velocities diagram. These flow boundaries were also compared with those available in the literature.