Deep Learning Based Volume Fraction Estimation for Two-Phase Water-Containing Flows

The estimation of water volume fraction in water-containing multiphase flows is important for several industrial applications. There are many published works detailing with methods to estimate and monitor water-containing multiphase flows based on different sensor modalities, including electrical capacitance tomography (ECT). A recently developed method based on ECT sensors to estimate the water volume fraction in water-containing two-phase flows utilizes the Hanai's mixture formula. Fresh or tap water is conductive and has large electrical permittivity, thus making the ECT-based estimation problem highly nonlinear and challenging to solve. Most of these works assume that the mixture state of water in the two-phase flow is known as either continuous or disperse. However, in practice, the state of water might not be known a priori. In this work, we propose a deep learning based approach to classify water-containing flows into water-dispersed or water-continuous flows and to estimate the water volume fraction present in the flow.