About the project
This project pioneers deep learning for turbulence modeling, focusing on wall-bounded flows. By combining convolutional neural networks (CNNs), generative adversarial networks (GANs), and physics-informed methods, it aims to develop hybrid predictive models that overcome current limitations. The research supports scalable, accurate simulations of multi-scale phenomena, advancing computational design across energy, transport, and biomedical applications.
This project explores the cutting edge of artificial intelligence for turbulence modelling in turbulent flows, particularly in wall-turbulent flows. Using advanced deep neural networks, this research aims to create hybrid predictive models that combine theoretical insights with data-driven approaches that reliably simulate turbulent flows. While CNNs and GANs have shown promise in capturing complex flow structures, such as reverse energy cascade, they also come with challenges, including the need for extensive, high-resolution training data and limitations in generalizing to new conditions. The goal is to develop predictive hybrid models for the computational design.
This project will push the boundaries of current methods, addressing specific limitations: CNNs’ difficulty with high-frequency flow features and GANs’ computational demands. The research will pioneer the use of physics-informed loss functions and a hybrid framework, enhancing model accuracy and applicability. The ultimate goal is to contribute models that can guide the design and development of fluid systems by predicting multi-physics, multi-scale phenomena, critical to advancing technologies in any field, such as sustainable energy, transport or medical.
Key project tasks include: creating specialized training datasets, optimizing network architectures for efficiency, exploiting the inclusion of physical constraints, integrating into computational fluid simulation (CFS) solvers, validating models in realistic settings. This work will also focus on scaling solutions to exascale computing, bringing new possibilities to turbulence simulations.
The project is part of a research collaboration with Professor De Angelis at the University of Bologna, contributing to a dynamic and interdisciplinary team at the forefront of AI and fluid mechanics.
Training will be provided on High-Performance Computing (HPC), programming, machine learning, and fluid dynamics. It will be realized through participation in summer schools and European training events, such as Cypher COST.
