Abstract:Machine learning approaches have been proven to be effective and efficient in mediating flow dynamics so as to realize drag reduction, vibration suppression, etc. In this article, an intelligent active flow control approach is proposed to enhance heat transfer of a heated circular cylinder at Re=100 undergoing vortex induced vibration in the lock-in condition. A high-fidelity solver based on the generalized interpolation-supplemented lattice Boltzmann method (GILBM) with moving reference frame (MRF) is used to establish the real-time numerical environment with structural motion, momentum transport, and heat transport. A reinforcement learning (RL) agent realized by proximal policy optimization (PPO) is adopted to establish the model-free controller, which receives sensorymotor cues as feedback signals and determines the real-time velocity issued from the jet as actuation. Through a group of trainings, each of whom involves 200 independent cases, the RL agent eventually learns an effective and efficient control strategy that improves the average heat transfer rate by 76.7%, quantified by the dimensionless Nusselt number. Through analyzing the spatiotemporal variations of vorticity and temperature fields, local Nusselt number and pressure coefficient along the cylinder surface, as well as the statistical flow quantities, it is proved that that the enhanced heat transfer is contributed by both the mediated vortex formation and shedding in vicinity of the blowing jet, as well as the augmented vibration. Extra tests also confirm that the synthetic jets play a much more significant role than the augmented vibration. Furthermore, the success of the control strategy is attributed mainly by the phase lag between the maximum vibration displacement and the moment when the synthetic jets oppose their directions. Moreover, the more realistic but more complicated case with the non-Oberbeck-Boussinesq (NOB) approximation is also tested, confirming good self-adaptivity of the learnt strategy. Through this study, it is verified that the RL to be a powerful tool for mediating flows involving complicated momentum and heat transport.

Machine learning open-loop control of a mixing layer

Towards human-interpretable, automated learning of feedback control for the mixing layer

Stabilization of a multi-frequency open cavity flow with gradient-enriched machine learning control

Iterative learning control of inhomogeneous distributed parameter systems - frequency domain design and analysis.

Jet mixing enhancement using machine learning control

Jet mixing optimization using machine learning control

Using CFD and Machine Learning to Explore Chaotic Mixing in Laminar Diameter-Transformed Stirred Tanks

Active Control of Vortex-Induced Vibration of a Circular Cylinder Using Machine Learning

Enhancing Heat Transfer from a Circular Cylinder Undergoing Vortex Induced Vibration Based on Reinforcement Learning

Application of multivariable feedback linearization for paper machine pressurized head box control

Variable structure control of unstable combustion

A Method of Active Control Nonlinear Combustion Instability

Modeling and Control of a Continuous Stirred Tank Reactor Based on a Mixed Logical Dynamical Model

Control of quasi-equilibrium state of annular flow through reinforcement learning

Data-driven Nonlinear K-L Turbulent Mixing Model Via Gene Expression Programming Method

Linear genetic programming control for strongly nonlinear dynamics with frequency crosstalk

Open-loop control of noise amplification in a separated boundary layer flow

From Active Learning to Deep Reinforcement Learning: Intelligent Active Flow Control in Suppressing Vortex-Induced Vibration

Optimal control for mixing enhancement in boundary layers at membrane walls

Active Control of Combustion Instabilities Using Model-Based Controllers