This lecture was given by Prof. Bernd R. Noack from TU Berlin, Germany in the framework of the von Karman Lecture Series on Machine Learning for Fluid Mechanics organized by the von Karman Institute and the Université libre de Bruxelles in February 2020. He gives a tour-de-force through select powerful methods of machine learning for data analysis, dynamic modeling, model-based control and model-free control. Focus is placed on few Swiss army knife methods which have proven to solve a large variety of ow problems. Examples are proximity maps, manifold learning, proper orthothogonal decomposition, clustering, dynamic modeling, control theory methods as contrast to machine learning control. In two following chapters of this course, the mentioned machine learning approaches are detailed for reduced-order modeling and for turbulence control. All methods are applied to a classical, innocent looking benchmark: the oscillatory twodimensional incompressible wake behind a circular cylinder at Re = 100 without and with actuation. This example has the advantage of being visually accessible to interpretation and foreshaddows already key challenges and opportunities of machine learning.

свернуть