Real-Time Air–Water Volume Fraction Prediction Using Deep Learning and High-Speed Imaging

Accurate real-time measurement of gas volume fractions is essential for analyzing multiphase flows, especially in pipe systems where understanding gas–liquid dynamics is critical for optimizing fluid transport and identifying potential issues like pressure surges or blockages. This study used a non-intrusive high-speed imaging to capture a transient water–air interactions within a transparent circular pipe, capturing a high-resolution visual data of a multiphase flow regimes. However, limited visual access in the turbulent flows with a gas–liquid interfaces restricts the visualization of interfacial dynamics. To address this, we developed a method to process video frames using a deep learning framework, combining the segment anything model (SAM) with the K-means clustering for an image segmentation. A Resnet-based neural network was trained to predict the water–air volume fractions across. This method enables real-time multiphase flow analysis with an application in the fluid transport optimization and a pipeline monitoring. The volume fraction correction was performed by comparing the predicted air–water volume fractions against low-intrusive WMS measurements, achieving an accuracy of 92.67%. This ensured accurate predictions and allowed for the assessment of uncertainties in the image processing methodology.