By Dr. Shahid Husain
Multiphase flow (MF) is the simultaneous flow of materials with two or more thermodynamic phases. MF occurs in numerous settings: bioengineering, conventional and nuclear power plants, oil and gas production and transport, pharmaceutical industry, combustion engines, chemical industry, flows inside the human body, biological industry, and process technology, to name a few. Researchers use experimental and theoretical techniques to study MF.
While advanced experimental techniques—such as magnetic resonance imaging, electrical impedance tomography, ultrasonic pulsed Doppler velocimetry, and neutron radiography—can provide precise spatial and temporal measurements of MF fields, they are costly and complex to implement. Alternatively, numerical models, including the Algebraic Slip model, Eulerian-Eulerian model, Eulerian-Lagrangian model, and Discrete Element methods, rely on different governing equations depending on the number of phases and numerical techniques employed. However, solving these equations often requires significant computational resources, making simulations expensive and time-consuming.
This project aims to develop an innovative machine learning (ML)-based hidden fluid dynamics approach for MF modelling. By integrating ML with fluid dynamics as shown in Fig. 1, using information derived from training data either generated through simulations or obtained from flow visualization images and governing equations, this method will significantly reduce the cost and time needed for MF analysis. The approach will enable more efficient design and optimization of MF systems, leading to lower operational and development costs while enhancing overall system performance.
Figure 1 Neural network diagram to develop ML-based model for MFs
These efforts align with the EU’s goals of improving energy efficiency, reducing emissions, and optimizing resource utilization in key sectors such as energy, manufacturing, and environmental engineering. By leveraging AI-driven models, researchers can enhance the accuracy and speed of simulations, leading to better predictions of complex fluid interactions in applications like carbon capture, hydrogen storage, and advanced manufacturing. These innovations support the EU’s broader objectives of digital transformation and climate neutrality.
We have begun our first study on bubbly flows, focusing on the injection of air bubbles into a water-filled tank as shown in Fig. 2. The problem is governed by Eulerian-Eulerian multifluid equations, and we generate training data for our machine learning algorithm using Ansys Fluent. So far, our preliminary results have been promising, and we are optimistic that our first model will be developed soon. This research has several practical applications, including improving efficiency in chemical reactors, optimizing aeration systems in wastewater treatment, and enhancing gas-liquid mixing in industrial processes. Additionally, accurate modelling of bubbly flows is crucial in energy systems such as nuclear reactors and carbon capture technologies.
Figure 2. Air Bubble rising in a water column
Shahid Husain is a Marie-Curie Postdoctoral Fellow at the Department of Mathematics and Statistics, University of Limerick and is active within MACSI (Mathematics Applications Consortium for Science and Industry). This project is funded by the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 101110330.