Blowing-only opposition control: Characteristics of turbulent drag reduction and implementation by deep learning

Opposition control is an effective active control strategy for drag reduction, which has been extensively investigated. In the current work, the essential characteristics of drag reduction by the blowing-only opposition control scheme (i.e., opposition blowing) in turbulent channel flow are investigated. It is demonstrated that, under the condition of constant wall-normal mass flux, the drag reduction achieved by the opposition blowing scheme is almost independent of the allocation of the blowing velocity among all the effective blowing points. This feature simplifies the complexity of the control scheme and provides great convenience for the application of the convolutional neural network (CNN) to implement the opposition blowing scheme, i.e., only the direction of the wall-normal velocity at the detection plane needs to be predicted. In this paper, both the streamwise and spanwise wall shear stresses are taken as the input of the CNN model, and the reasonability of the CNN model is verified from a statistical perspective. It is found that as long as the directions of the large wall-normal velocity fluctuations are accurately predicted, the opposition blowing scheme can be successfully implemented, in which the CNN model is able to ensure a high prediction accuracy. Furthermore, applying the trained network model to a flow at a higher Reynolds number than the training set can still accurately predict the directions of the large wall-normal velocity fluctuations, which generalizes the applicability of the CNN model.