Identification of Rising Gas-Liquid Two-Phase Flow Regimes Using Ultrasound with Machine Learning

Abstract

A method for identifying the flow regimes of gas-liquid two-phase flow using ultrasound was developed in this study. The method of analyzing the gas-liquid two-phase flow rising in a vertical pipe can be applied to the sequential measurement method in an airlift pump which is used for pumping rare earth resources existing in the deep sea bottom. The distributions of the ultrasound echo intensity and velocity calculated from Doppler frequencies are output as image data with the distance from the ultrasound transducer and profiling time as the two axes, and the images are processed using various filters. The results suggest that it is possible to distinguish between small and large bubbles by focusing on the reflection intensity obtained from the region behind the bubbles on the measurement line and the flow velocity after the passage of the bubbles. By identifying between small and large bubbles, the void fraction, which is an index of the flow regimes, was calculated from the echo data. On the other hand, it remains a challenge to set appropriate thresholds for various objects and environments and to evaluate them quantitatively. Thus, we also attempted to classify bubbly flow and slug flow by using machine learning on the image data.