Deep convolutional neural networks (CNN) have originally been designed for image processing task. 
However, their structure was also shown to be very suitable for modelling task in fluid dynamics.
In an inital study we developed WakeNet, a deep convolutional neural network predicting the 3D flow field behind wind turbines, trained with high-fidelity simulation data.
Results from the initial were very promising, but left various open questions.

 

This Master thesis shall build on the initial implementation of WakeNet, perform a more rigorous validation of the model and eventually implement and test new CNN architectures to improve the model's prediction capabilities (e.g., ResNet or UNet)

- a strong interest and background knowledge in fluid dynamics and wind turbine aerodynamics
- at least basic experience in python
- a strong interest in deep learning