Optimizing Skin-Friction Drag Reduction via Low-Amplitude Wall-Normal Blowing Techniques Using a Bayesian Optimization Framework

In this research, we used low-amplitude wall-normal blowing to reduce the skin-friction drag of zero-pressure-gradient turbulent boundary-layer flows in a wind tunnel, optimized using a novel Bayesian optimization framework, NUBO (Newcastle University Bayesian Optimization). NUBO was used to optimize control parameters including amplitude, frequency, and wavelength of actuation to identify control strategies which give high skin-friction drag reduction with a continual change in wind speed between 5 m/s and 20 m/s. A hot-wire probe was used to measure the streamwise velocity across the linear sublayer in the turbulent boundary layer to calculate skin friction coefficients. Comparisons to high-fidelity simulations matching the wind tunnel experiments were made to investigate the changes in flow physics during control. It is shown that NUBO can find various sets of parameters to achieve significant drag reduction corresponding to different fluid flow mechanisms. This highlights the potential of using machine learning approaches for flow control applications.