DeepTurb – Deep Learning in und von Turbulenz

The application of machine learning to experimental measurements and numerical simulation calculations of turbulent flows opens up unique opportunities to reclassify complex data according to physical criteria. Thus, a previously missing understanding of the fundamental transport processes can be gained for their more effective modeling. Artificial intelligence will be used to extract the dynamics of large-scale patterns of turbulence from extensive measurement and simulation data horizontally in extended turbulent convection flows and to predict them in highly simplified nonlinear dynamic models based on neural networks. These applications also require an extension of the mathematical foundations of machine learning, which make turbulence prediction possible in the first place.