2D SPATIAL/ 1D TEMPORAL VOID FRACTION ANALYSIS OF VERTICAL UPWELLING FLOW USING MULTIPLE CURRENT-VOLTAGE AND CONVOLUTIONAL NEURAL

Abstract

One of the essential parameters for the safety and optimal control of industrial facilities is the void fraction α which represents the percentage of gas phase volume in a unit volume of a pipeline. However, accurate α estimation is difficult due to the unsteady and inhomogeneous flows. In this study, a new concept of 2D spatial/1D temporal (3D) void fraction has been analyzed by a combination of multiple current-voltage and convolutional neural network (MCV-CNN). The MCV-CNN is composed of four steps which are 1) 3D in-situ void fraction measurement by wire-mesh sensor (WMs), 2) Simulated voltage by electrical simulation 3) Measured voltage by MCV system and 4) training parameter in CNN for accurate 3D estimated void fraction α^. In order to determine α^, epperiments on the transition zone from bubbly to slug flow were conducted in a vertical pipe. As a result, the MCV-CNN predicts the 3D α^ with an averaged normalized cross correlation of 0.686.