Deep learning architecture for sparse and noisy turbulent flow data

Dimitris Drikakis, Ioannis Kokkinakis, Filippos Sofos
University of Nicosia, Nicosia, Cyprus

The article “Navigating the Unpredictable: Deep Learning’s Role in Mastering Turbulent Flows” was promoted as “Featured” by the prestigious journal of the American Institute of Physics

The article “Deep learning architecture for sparse and noisy turbulent flow data” by Professor Drikakis, UNIC Vice President for Global Partnerships and Executive Director of Research & Innovation Office, and his team, Dr Ioannis Kokkinakis and Dr Filippos Sofos, published in the premier journal Physics of Fluids of the American Institute of Physics, was selected by the Editors as one of the journal’s best and was chosen to be promoted as a Featured Article.

Image: Contour plot of the vorticity magnitude illustrating the flow asymmetry, turbulent boundary and shear layers.

Objectives of the study

  • To better understand how handling sparse and noisy data accurately will affect the success of deep learning models in fluid dynamics applications
  • To investigate how to remove noise from flow images after training the model with high-resolution and noisy images
  • To provide a model as an alternative when repetitive experiments are complex and only a small amount of noisy data is available.

What Does This Mean?

  • 1
    The success of deep learning models in fluid dynamics applications will depend on their ability to handle sparse and noisy data accurately.
  • 2
    Deep learning models can be developed for reconstructing turbulent flow images from low-resolution counterparts encompassing noise
  • 3
    It is possible to remove noise from flow images after training a deep learning model with high-resolution and noisy images
  • 4
    Deep learning model training could function as an alternative in fluid dynamics, when repetitive experiments are complex and only a small amount of noisy data is available.

Real-World Applications

This model, rooted in the principles of artificial intelligence, specifically employs a convolutional neural network (CNN) to transform low-resolution and noisy images into clear, high-resolution depictions of turbulent flows. Such an advancement is not just academic; it has real-world applications that touch on various aspects of daily life and global challenges.

Aircraft Design and Safety: In the aerospace industry, understanding turbulent flows is key to designing safer and more efficient aircraft. Turbulent air can cause uncomfortable rides or even dangerous situations. The researchers’ model can help engineers advance the study of airflows around aircraft wings at a fraction of the current time and cost, leading to designs that minimize turbulence’s impact on flights.

Climate Science and Weather Prediction: Climate scientists can use this model to better understand the turbulent dynamics of the atmosphere and oceans. For instance, accurately reconstructing the turbulent flows in hurricane formation can improve prediction models, giving communities more time to prepare for severe weather events. Similarly, understanding ocean currents can help in modeling climate change scenarios with greater accuracy.

Medical Applications: In medicine, turbulent blood flow can indicate cardiovascular issues. The ability to reconstruct and analyze blood flow patterns from imperfect data could enhance diagnostic imaging techniques, such as MRI or ultrasound, making them more precise and leading to early detection of life-threatening conditions.

Renewable Energy: For the renewable energy sector, particularly wind farms, understanding and optimizing the interaction between wind and turbines is essential for efficiency. The model’s ability to reconstruct turbulent flows can lead to better placement of turbines and designs that extract more energy from the wind, accelerating our transition to green energy.

Environmental Conservation: Turbulent flows in rivers and oceans are crucial for ecosystems, affecting everything from nutrient distribution to the migration patterns of aquatic life. By providing a tool to study these flows in detail, the model can aid in conservation efforts, helping to preserve biodiversity and maintain healthy ecosystems.

Media and Press Materials

Contact Information

For more information about this study, please contact Professor Dimitris Drikakis, UNIC Vice President for Global Partnerships and Executive Director of Research & Innovation Office, at [email protected].