Workgroup Flow Modelling and Control

Prof. Dr. Bernd Noack and Dr. Richard Semaan

The flow modelling and control workgroup researches and develops tools for flow state estimation and for closed-loop flow control applications. We address key challenges of turbulence control from the inherent nonlinearity of actuation mechanism to noise in experiments.

Research activities within the flow modelling and control group

Current research of the flow prediction and control workgroup at the institute of fluid mechanics focuses mainly on 3 fields:

Model-based control: The first step for model-based control requires the development of a reduced order model (ROM). Due to their low complexity, ROMs enable real-time flow control via fast sensing and actuation. Their relevance also relates to their ability to capture the important flow physics, while being simple enough for online control. Research activities include:

  • Development of reduced-order models e.g.: POD-Galerkin, CROM, DMD …

  • Model calibration techniques e.g.: regularized least-squares, 4D-Var...

  • Optimal sensor placement e.g.: stochastic estimation, machine learning …

Model-free control: The model-free control identifies the optimal nonlinear control law in an automatic (unsupervised) self-learning manner. This approach employs genetic programming and is termed Genetic Programming Control (GPC) or Machine Learning Control (MLC). Research activities include:

  • Further development of GPC algorithm.

  • GPC coupling with ROM

  • Experimental testing in wind tunnels

Data Mining: Data mining or machine learning is a rapidly evolving field widely used in the scientific as well as the business community such as the financial sector and internet search engines. Machine learning in fluid dynamics is a new and novel concept, which holds big promises.

Flow Viz. of an airfoil with a blown Coanda flap


  last changed 20.01.2016
TU_Icon_E_Mail_1_17x17_RGB pagetop