Study on Liutex-driven reward for intelligent flow control by dynamic feature-based deep reinforcement learning
Source abstract
Abstract
The issue of flow separation over an airfoil under weak turbulent conditions is addressed and resolved through the deep reinforcement learning (DRL) strategy. To suppress the generation of separation flow and the instability of the vortex shedding alley over an airfoil, a novel reward function named RLiutex, considering the suppression of the rigid rotation intensity and core size of the vortex, is first proposed in this paper. The great potential of this Liutex-driven reward for effective large eddy elimination and aerodynamic optimization is verified in this work. Furthermore, a dynamic feature-based DRL (DF-DRL) framework is redeveloped to markedly enhance the learning efficiency and convergence speed. The combination of the Liutex-driven reward function with the DF-DRL framework results in an exceptional aerodynamic performance, with minimal fluctuations in both drag and lift coefficients, highlighting the potential of this approach for advanced flow control strategies.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Model Reduction and Neural Networks · Real-time simulation and control systems · Fluid Dynamics and Turbulent Flows
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Energy and Power Engineering, University of Shanghai for Science and Technology 1 , Shanghai 200093,
Shanghai Key Laboratory of Multiphase Flow and Heat Transfer for Power Engineering 2 , Shanghai 200093,
Funding
National Natural Science Foundation of China · 2021 Shanghai Chenguang Program
References (31)
Show source-provided references
1. Robust and adaptive deep reinforcement learning for enhancing flow control around a square cylinder with varying Reynolds numbers10.1063/5.0207879 ↗
2. 2017 · 49
3. Flow control in wings and discovery of novel approaches via deep reinforcement learning10.3390/fluids7020062 ↗
4. Machine learning in aerodynamic shape optimization10.1016/j.paerosci.2022.100849 ↗
5. Active control of vortex-induced vibration of a circular cylinder using machine learning10.1063/1.5115258 ↗
6. Recent advances in applying deep reinforcement learning for flow control: Perspectives and future directions10.1063/5.0143913 ↗
7. Artificial neural networks trained through deep reinforcement learning discover control strategies for active flow control10.1017/jfm.2019.62 ↗
8. Efficient deep reinforcement learning strategies for active flow control based on physics-informed neural networks10.1063/5.0213256 ↗
9. 2019
10. T. P. Lillicrap , J. J.Hunt, A.Pritzel, “Continuous control with deep reinforcement learning,” arXiv:1509.02971 (2015).
11. J. Schulman , F.Wolski, P.Dhariwal, “Proximal policy optimization algorithms,” arXiv:1707.06347 (2017).
12. Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor
13. DRLinFluids: An open-source Python platform of coupling deep reinforcement learning and OpenFOAM10.1063/5.0103113 ↗
14. Investigation of vortex motion mechanism of synthetic jet in a cross flow10.1063/5.0086084 ↗
15. The mechanism of turbulent drag reduction with wall oscillation10.1016/s0142-727x(00)00070-9 ↗
16. Direct shape optimization through deep reinforcement learning10.1016/j.jcp.2020.110080 ↗
17. Optimization of configuration of corrugated airfoil using deep reinforcement learning and transfer learning10.1063/5.0134198 ↗
18. Reinforcement-learning-based control of convectively unstable flows10.1017/jfm.2022.1020 ↗
19. Robust flow control and optimal sensor placement using deep reinforcement learning10.1017/jfm.2020.1170 ↗
20. Reinforcement-learning-based control of confined cylinder wakes with stability analyses10.1017/jfm.2021.1045 ↗
21. An application of data driven reward of deep reinforcement learning by dynamic mode decomposition in active flow control10.1007/978-3-031-44947-5_92 ↗
22. Dynamic feature-based deep reinforcement learning for flow control of circular cylinder with sparse surface pressure sensing10.1017/jfm.2024.333 ↗
23. Deep reinforcement learning based synthetic jet control on disturbed flow over airfoil10.1063/5.0080922 ↗
24. Active control of transonic airfoil flutter using synthetic jets through deep reinforcement learning10.1063/5.0232492 ↗
25. Deep reinforcement learning for turbulent drag reduction in channel flows10.1140/epje/s10189-023-00285-8 ↗
26. Applying deep reinforcement learning to active flow control in weakly turbulent conditions10.1063/5.0037371 ↗
27. Deep reinforcement learning for flow control exploits different physics for increasing Reynolds number regimes10.3390/act11120359 ↗
28. Data-efficient deep reinforcement learning with expert demonstration for active flow control10.1063/5.0120285 ↗
29. A novel framework for predicting active flow control by combining deep reinforcement learning and masked deep neural network10.1063/5.0194264 ↗
30. Rortex—A new vortex vector definition and vorticity tensor and vector decompositions10.1063/1.5023001 ↗
31. Explicit formula for the Liutex vector and physical meaning of vorticity based on the Liutex-shear decomposition10.1007/s42241-019-0032-2 ↗