Policy-based optimization for drag reduction via spanwise wall oscillations
摘要
This study introduces a novel computational framework that synergistically integrates policy-based optimization (PBO) and implicit large eddy simulations (LES) to optimize oscillating spanwise wall motions for drag reduction in turbulent channel flows. Drag reduction remains a critical challenge in fluid dynamics, particularly in the context of improving energy efficiency in various engineering systems. Previous investigations aiming to determine optimal spanwise wall oscillation (SWO) parameters have predominantly relied on computationally expensive methods, such as parametric grid search using direct numerical simulations. Consequently, these studies have been limited in their exploration of the parameter space due to the associated high computational costs. The primary contribution of the present study lies in presenting a significant advancement through a more efficient and cost-effective exploration of the SWO parameter space, without compromising the precision of the simulations. By leveraging machine learning techniques, specifically PBO, the proposed framework enables rapid convergence to optimal control strategies while minimizing the number of simulations required. This approach marks a departure from the conventional a priori parameter selection based on grid search methods. Beyond this, the study also demonstrates the effectiveness of PBO for parametric optimization in turbulent flow control applications and corroborates the previously identified optimal drag reduction parameters while providing deeper insights into their precise optimal ranges and sensitivities. The optimization process is conducted in two stages: initially focusing solely on maximizing drag reduction, followed by the incorporation of control costs to optimize the net energy balance of the system. The optimal oscillation parameters obtained for drag reduction closely align with those reported in the literature, with the optimal period