Real-time feedback stall control of an airfoil at large Reynolds numbers using linear genetic programming

Machine learning control is applied in real-time to an airfoil equipped with variable-velocity jets and pressure sensors in a closed-loop wind tunnel. The objective of the control is to improve the lift-to-drag ratio using an array of variable velocity jets located at the leading edge of the wing mo...

Full description

Saved in:
Bibliographic Details
Published in:Physics of fluids (1994) 2022-04, Vol.34 (4)
Main Authors: Passaggia, Pierre-Yves, Quansah, Asamaning, Mazellier, Nicolas, Maceda, Guy Y. Cornejo, Kourta, Azeddine
Format: Article
Language:English
Subjects:
Actuation
Actuators
Airfoils
Angle of attack
Control methods
Control systems
Control theory
Drag
Engineering Sciences
Feedback control
Fluid dynamics
Fluid flow
Fluids mechanics
Genetic algorithms
High Reynolds number
Jets
Leading edges
Lift
Machine learning
Mechanics
Physics
Pressure sensors
Real time
Stalling
Wind tunnels
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Machine learning control is applied in real-time to an airfoil equipped with variable-velocity jets and pressure sensors in a closed-loop wind tunnel. The objective of the control is to improve the lift-to-drag ratio using an array of variable velocity jets located at the leading edge of the wing model. Pressure sensors are located along the chord, at mid-span, and are used in a feedback control strategy where the symbolic control laws are optimized using a linear genetic programming control (LGPC) algorithm. For angles of attack at the onset of stall, we show that the best control law is able to outperform the best open-loop control strategy in the case of a single-input multiple-output control loop. Results are reported for Reynolds numbers at both half a million and one million. In particular, the lift-to-drag ratio is improved by 4% compared to the best open-loop strategy, which corresponds to a relative increase by a factor 3. In the case of leading-edge separation conditions, LGPC builds a control law performing similarly to the best open-loop strategy while minimizing the actuation power. This study suggests that in order to improve the control authority, the LGPC strategy is sufficiently mature. This control methodology highlights that in order to improve the aerodynamic performances, novel fluidic actuators providing access to the orientation of the jets could enable breakthroughs for high-Reynolds-number experimental demonstrators.
ISSN:1070-6631
1089-7666
DOI:10.1063/5.0087874