Prediction and understanding of the wall-shear stress modulation by non-linear interactions based on novel machine learning techniques

Abstract

This project addresses a critical challenge in fluid dynamics: the accurate measurement and understanding of wall-shear stress in turbulent flows. Wall-shear stress, a measure of friction drag and dynamic loads, is crucial for numerous engineering applications, including the design of aircraft, trains, and chemical processes, as well as advancements in wind power generation and biomedicine. However, its instantaneous and spatially resolved measurement in experimental settings remains exceptionally difficult due to technical limitations. The project proposes a novel approach by leveraging modern deep learning algorithms to predict the wall-shear stress based on more easily accessible velocity measurements such as those obtained from Particle-Image Velocimetry (PIV) experiments. This approach goes far beyond traditional methods that often rely on linear assumptions or provide only single-point, temporal data, limiting a comprehensive understanding of complex, non-linear, and multi-scale interactions within turbulent flows. A key aspect of this research is the development of neural network architectures specifically designed to learn a mapping function from two-dimensional velocity fields located in the outer layer of a turbulent flow - which are usually easily measured - to the instantaneous wall-shear stress distribution. This will not only provide novel insight into the non-linear interactions that modulate the wall-shear stress but also make the complex physics understandable to humans, which is vital for developing more efficient and sustainable technologies. The relevance of these findings to the interested public is profound. Improved prediction and understanding of wall-shear stress can lead to more efficient transportation, as reducing drag on aircraft and trains translates directly into lower CO2 emissions and decreased fuel consumption. In human medicine, this research holds the potential for improved, personalized prediction and treatment of cardiovascular diseases that are significantly affected by wall-shear stress. Moreover, the advancement of renewable energy technologies, such as more efficient wind turbine designs, can be achieved through a deeper understanding of the forces exerted on their blades. By bridging the gap between complex fluid physics and interpretable, data-driven models, this project delivers foundational knowledge and practical tools for a wide range of applications that impact daily life and global sustainability.