Anina Šarkić: Machine Learning in Wind Engineering

Machine Learning Seminar presentation

Topic: Machine Learning in Wind Engineering

Speaker: Dr. Anina Šarkić, University of Luxembourg, Department of Computational Science

Time: Wednesday 2020.11.25, 10:00 CET

How to join: Please contact Jakub Lengiewicz


With the development of construction technology and materials, more light and super-flexible structures have been built all over the world. As a consequence, they are becoming very sensitive to the wind loads and, in addition, more complex wind flow patterns are developed. In design phase, this more detailed wind loads and flow information comes from wind tunnel tests. However, they are very costly when several geometrical scenarios are to be analyzed. In addition, wind tunnels also do not naturally allow the monitoring of important quantities of interest over large control volumes. Another approach that can overcome some of the drawbacks of wind tunnel methodology relays on computational fluid dynamics, yet cannot be used in isolation from wind tunnels.

Therefore, this presentation explores other effective ways that may go beyond existing methodologies that rely on machine learning. In particular, prediction of wind loads on the high-rise building using backpropagation neural network combined with POD will be shown.

Additional material:

[1] Dongmei, H., Shiqing, H., Xuhui, H., Xue, Z., (2017), Prediction of the wind loads on high-rise buildings using BP neural network combined with POD, Journal of wind engineering and industrial aerodynamics, 170, 1-17. (

[2] Tian, J., Gurley, K., R., Diaz., M., T., Fernandez-Caban, P., L., Masters, F., J., Fang, R., (2020), Low-rise gable roof buildings pressure prediction using deep neural networks, Journal of wind engineering and industrial aerodynamics, 196 (

[3] Bernardini, E., Spence, S., M., J., Wie, D., Kareem, A., (2015), Aerodinamic shape optimization of civil structures: A CFD-enabled Kringing-based approach, Journal of wind engineering and industrial aerodynamics, 144, 154-164. (