Machine-Learning Based Active Flow Control
Abstract：In this talk, some recent applications of machine learning (ML) in active flow control (AFC) will be introduced. Here the term AFC means that the control is realized by injecting a small amount of energy into existing flow systems. Compared to its counterpart, i.e., passive flow control, AFC is adaptive and on-demand, and hence has a much wider operating range. First, the use of generic-programming (GP) selected explicit control laws for the blowing/suction enabled control of vortex-induced vibration (VIV) of a circular cylinder will be presented. Second, the use of deep reinforcement learning (DRL) for eliminating the velocity deficit behind a circular cylinder using a group of windward-suction-leeward-blowing (WSLB) actuators will be introduced. Last, the use of DRL for finding best drag reduction strategies for a fixed circular cylinder will be demonstrated at different flow conditions. Through these ML based AFC studies, some new and unexpected control strategies have been revealed.
Bio-sketch：Dr. Hui Tang is an Associate Professor, Director of Research Center for Fluid-Structure Interactions, and Associate Head of Department of Mechanical Engineering, The Hong Kong Polytechnic University. He received his BEng and MEng degrees from Tsinghua University, and his PhD degree in Aeronautical Engineering from University of Manchester. Prior to joining HK PolyU, he worked in Nanyang Technological University, and University of Michigan - Ann Arbor. His research interests include aerodynamics/hydrodynamics, active flow control, fluid-structure interaction, and heat and mass transfer. In these fields he has published more than 90 journal/conference papers. He is/has been a member of scientific or organizing committees for a number of international conferences/symposiums. He is now serving as an Executing Committee Member of Hong Kong Society of Theoretical and Applied Mechanics (HKSTAM).