Active Flow Controller and Control Method Based on Deep Reinforcement Learning

Abstract

The invention provides an active flow controller and a control method for suppressing vortex-induced vibration based on deep reinforcement learning. The active flow controller of the present invention establishes a deep reinforcement learning decision-making agent based on the Soft Actor-Critic algorithm, and outputs a robust real-time control strategy by continuously interacting with the flow environment. By establishing a reward and punishment function related to the vortex-induced vibration state in the cross-flow direction of the cylinder and the surface resistance of the cylinder, dynamically learn and adjust the weight of the artificial neural network in the decision-making agent, and establish a mapping relationship from flow conditions such as flow environment speed and pressure to control actions , so that an active flow controller is obtained. By using the active flow controller of the present invention to control the suction and blowing devices installed symmetrically on the cross-flow poles of the cylinder, the two control objectives of vibration suppression and drag reduction of the cylinder can be realized.

Description

technical field [0001] The invention relates to an active flow control method based on deep reinforcement learning, which dynamically adjusts the parameters of the artificial neural network through the continuous interaction between the decision-making agent and the flow field environment, and controls the lateral flow of the cylinder to the pole according to the flow state observation of the numerical simulation environment The blowing and suction device affects the shedding of the vortex on the surface of the cylinder and the fluid-solid coupling process, thereby achieving the effects of suppressing vortex-induced vibration and reducing drag, and belongs to the field of active flow control. Background technique [0002] Designing active flow control strategies is a tedious task. Compared with passive flow control, the actuating mechanism of active flow control is often more complicated. Therefore, when designing an active flow control strategy, designers also need to desi...

AI Technical Summary

Problems solved by technology

Compared with traditional wind tunnel experiments, the emergence of computational fluid dynamics technology has provided convenience for the exploration of control strategies, but due to the high dimensionality and strong nonlinear characteristics of fluid mechanics, the exploration of strategies requires a large amount of computing costs. Active flow control strategies in most studies are limited to simple constant or harmonic inputs, therefore, there is a need to develop an effective control strategy exploration method for active flow control mechanisms that takes full advantage of the control possibilities of active control

Images