Somerville’s Tutorial Fellow in Engineering, Professor Konstantina Vogiatzaki, and her postdoc Dr Giovanni Tretola have won best paper award at the recent International Conference on Liquid Atomization and Spray Systems (ILASS), held in Lisbon, Portugal
Postdoctoral Research Assistant Dr Giovanni Tretola, the paper’s first author, and Professor Konstantina Vogiatzaki were recognised for their paper, A Mixture Density Network Model for Droplet Velocity Statistics in Jet-in-Crossflow Atomisation.
Professor Vogiatzaki and Dr Tretola (centre)
The research explores how machine learning can help predict the behaviour of tiny liquid droplets formed when a jet of liquid enters a moving stream of gas. Understanding how these droplets move is important for applications including fuel injection, combustion and other spray technologies.
Using data from detailed computer simulations, the researchers trained a machine-learning model to predict the distribution of droplet speeds under different operating conditions. Rather than estimating just an average speed, the model captures the wider range of velocities and provides information about the uncertainty in its predictions. The results show that the model can reproduce key patterns in the simulated data, including changes in the distribution of droplet speeds. Once trained, it can estimate these distributions much faster than running the detailed simulations, within the range of conditions studied.
The findings demonstrate how machine learning can complement traditional computational methods, offering new opportunities to study sprays and the complex interactions between liquids and gases.
Commenting on the award, Dr Tretola said: “I am happy that this work has been recognised. It’s encouraging to see how machine learning can complement other approaches to studying sprays and multiphase flows, and contribute to research in this field.”
The Oxford research group also contributed five other presentations at the conference, showcasing the breadth of its work in spray and multiphase-flow research.
Professor Konstantina Vogiatzaki reflected on the best paper award, saying that, “Over the past few years, our research group has been increasingly exploring the potential of machine learning to better understand and predict complex fluid flows. This work represents another important piece of that larger puzzle, and it is particularly rewarding to see it recognised by the international research community. What excites me most is the potential of machine learning to go beyond simply accelerating predictions, helping us uncover the underlying physics of complex flow phenomena. In particular, it offers exciting opportunities to gain insight into microscopic fluid dynamics processes that are extremely challenging, or sometimes even impossible, to resolve experimentally.”
The International Conference on Liquid Atomization and Spray Systems (ILASS) brings together researchers and specialists working on the formation, movement and application of sprays. The conference provides an opportunity to share new findings and discuss developments in the science and technology of liquid atomisation.