Numerical Simulation on Drag Reduction for Flow Past Two Tandem Cylinders with Reinforcement Learning

Drag reduction for flows over blunt bodies has always been a hot topic of research in aerospace and other fields, which is very important to improve aircraft performance and reduce energy consumption. In this paper, the drag reduction of flow past two tandem cylinders was studied by placing zero-mass jets on the cylinder surface and using a deep reinforcement learning method to control the jet flow adaptively. Firstly, the numerical simulation of the flow past two cylinders with equal diameters was carried out, and the resulting lift and drag coefficient were in good agreement with the literature, which verified the grid independence of the numerical method. Then, a series of two cylinders with L/D=2 at different jet placement angles were studied to explore the drag reduction effect of the cylinders at Re=200. The results show that when the jet placement angle is 90° and 270°, the drag coefficient of the front cylinder and the rear cylinder after training is reduced by 1.16%, 9.9%, respectively, and the overall drag coefficient of the two cylinders is reduced by 3.7%. When the jet placement angle is 60° and 300°, the drag coefficient of the front cylinder is reduced by 0.77%, the drag coefficient of the rear cylinder is reduced by 3.35%, and the overall drag coefficient of the two cylinders is reduced by 1.75%. Then, the drag reduction effect of two cylinders with unequal diameters with L/D=6 at 90° and 270° was studied when Re=100. After training, the drag coefficient of the front and rear cylinders is reduced by 18.9% and 63.2%, respectively, and the overall drag coefficient is reduced by 33.6%. This study can provide reference for drag reduction technology of blunt body using zero-mass jet.