W czasopiśmie "Signal, Image and Video Processing" (Springer, IF:2.1) ukazał się artykuł będący efektem współpracy międzynarodowej KMiSD: Mayet A.M., Mohammed S.A., Hanus R., Guerrero J.W.G., Qamar S., Loukil H., Shukla N.K., Kowalczyk A.: Fluid-independent volume fraction measurement in liquid-gas flow using tomographic data and deep neural networks. Signal, Image and Video Processing 20 (2026), 337. (https://link.springer.com/article/10.1007/s11760-026-05397-0)
Abstract:
Accurately assessing the volume fraction in multiphase flow systems is important for a wide range of engineering applications, such as power plants and the petroleum industry. In this paper, a method is introduced that uniquely combines sinograms—a data representation originally used in computed tomography to capture projections from multiple angles, but in this case constructed from measurements across different electrode pairs—and Artificial Neural Networks (ANN) to measure volume fractions independently of the liquid phase type, achieving high accuracy across different liquids. Using COMSOL Multiphysics, simulations were conducted to generate data for various liquid-gas combinations, including water, gasoline, gasoil, crude oil, and oil. An 8-electrode concave capacitive sensor was employed to capture capacitance values, which were then converted into sinograms. These sinograms are used to be the input values for a Neural Network model created in Python, enabling precise volume fraction estimations regardless of the liquid phase involved. Additionally, tomographic images of the pipe were created, which showed acceptable results only for water. This method demonstrates versatility and accuracy in measuring volume fractions for various liquid phases, highlighting its potential for a wide industrial application. It should be noted that the proposed technique has been assessed only for perfectly stratified flows, and its applicability to other flow regimes has not yet been verified.
