The modelling of turbulent flows is recognized as a challenging problem in science. To this day, the complexity of fluid flows is still not fully understood. The effects of turbulent flows have to be considered in a great many areas, when building an aircraft, car or a prosthetic heart valve and predicting the weather, for example.
Researcher at the Swiss Federal Institute of Technology in Zurich (ETH) have now developed new reinforcement learning (RL) algorithms and combined them with physical insight to model turbulence. ETH pioneered the interfacing of AI and turbulent flows 25 years ago, Petros Koumoutsakos, Professor at the Laboratory for Computational Science and Engineering at ETH Zurich, explained in a press release. However, computers back then were not powerful enough to test many of the ideas. This all changed with the “Piz Daint” supercomputer at the Swiss National Supercomputing Centre (CSCS) in the canton of Ticino. It allowed researchers to explore their idea and develop it successfully.
Koumoutsakos and his team believes that their newly developed method will not only be of importance in the construction of cars and in weather forecasting. “For most challenging problems in science and technology, we can only solve the ‘big scales’ and model the 'fine' ones,” says Koumoutsakos. “The newly developed methodology offers a new and powerful way to automate multiscale modelling and advance science through a judicious use of AI.”