New Deep Learning Approach Enhances Flow Control for Engineering Applications

Recent research has introduced a novel approach to controlling the wake flow behind a confined square cylinder using a combination of deep learning and reinforcement learning techniques. The study, titled "Efficient Active Flow Control Strategy for Confined Square Cylinder Wake Using Deep Learning-Based Surrogate Model and Reinforcement Learning," was conducted by Meng Zhang, Mustafa Z. Yousif, Minze Xu, Haifeng Zhou, Linqi Yu, and HeeChang Lim.

The researchers developed a deep learning model-based reinforcement learning (DL-MBRL) framework that integrates a deep learning surrogate model (DL-SM) with computational fluid dynamics (CFD) simulations. This innovative approach aims to suppress wake vortex shedding, a phenomenon that can lead to inefficiencies in various engineering applications. By alternating between the DL-SM and CFD simulations, the framework significantly reduces computational costs associated with traditional methods.

The DL-SM employs a combination of a Transformer model and a multiscale enhanced super-resolution generative adversarial network (MS-ESRGAN) to accurately model complex flow dynamics. The model was trained on two-dimensional direct numerical simulation (DNS) data, demonstrating strong agreement with DNS results, which validates its accuracy.

One of the key findings of the study is that the DL-MBRL framework reduced training time by approximately 49.2%, decreasing from 41.87 hours to 20.62 hours. Both the DL-MBRL and a model-free reinforcement learning (MFRL) approach achieved significant reductions in shedding energy—98%—and in the standard deviation of the lift coefficient (C_L) by 95%. However, the DL-MBRL approach showed improved exploration capabilities, addressing issues related to nonzero mean lift coefficients that were present in the MFRL method.

This research highlights the potential of combining deep reinforcement learning with deep learning surrogate models for enhanced active flow control, which could have significant implications for various fields, including aerospace, automotive, and civil engineering. The full paper is available for further reading at arXiv:2408.14232.