Surrogate‐based optimization for active drag reduction of turbulent boundary layer flows

Abstract

AbstractTwo surrogate‐based optimization strategies using support vector regression (SVR) and Gaussian process regression (GPR) as surrogates are investigated to guide the design of actuation parameters for active drag reduction techniques in turbulent boundary layer flows encountered at civil airplanes in cruise flight and high‐speed trains. As an approximation, the turbulent flow over a flat plate subjected to spanwise traveling transversal sinusoidal surface waves is simulated by wall‐resolved large‐eddy simulations (LESs). These simulation data are used to model the dependence of the objective drag reduction on the actuation parameters, that is, the optimization variables. In this work, the previous purely exploitative approach of SVR‐based ridgeline optimization is extended to GPR‐based Bayesian optimization to further automate the simulation‐driven tuning of the actuation parameters.