Opposition control of turbulent channel flow with wall pressure using deep neural network

The opposition feedback control proposed by Choi et al. (JFM, 1994) successfully reduces the skin-friction of wall-bounded turbulent flow by using blowing/suction at the wall in opposition to the near wall velocity. The motivation of this study is to explore the possibility of using the wall pressure as a control input for the opposition control instead of the near wall velocity. For this purpose, we build a deep neural network (DNN) to predict the near wall velocity from the information of wall pressure in turbulent channel flow. For the learning process to build DNN, instantaneous flow data sets are obtained from direct numerical simulation of turbulent channel flow at Re_τ = 178. We examine the prediction performance of DNN according to the plane size of wall pressure as an input, type of DNN model, number of learning data sets, and etc. It is found that the near wall velocity can be successfully predicted by DNN with the wall pressure input. Also, we conduct opposition control of turbulent channel flow based on the DNN constructed, and it is shown that its control performance of skin-friction reduction is similar to that of the original opposition control.