This paper focuses on the active flow control of a computational fluid dynamics simulation over a range of Reynolds numbers using deep reinforcement learning (DRL). More precisely, the proximal policy optimization (PPO) method is used to control the mass flow rate of four synthetic jets symmetrically located on the upper and lower sides of a cylinder immersed in a two-dimensional flow domain. The learning environment supports four flow configurations with Reynolds numbers 100, 200, 300, and 400, respectively. A new smoothing interpolation function is proposed to help the PPO algorithm learn to set continuous actions, which is of great importance to effectively suppress problematic jumps in lift and allow a better convergence for the training process. It is shown that the DRL controller is able to significantly reduce the lift and drag fluctuations and actively reduce the drag by ∼5.7%, 21.6%, 32.7%, and 38.7%, at Re = 100, 200, 300, and 400, respectively. More importantly, it can also effectively reduce drag for any previously unseen value of the Reynolds number between 60 and 400. This highlights the generalization ability of deep neural networks and is an important milestone toward the development of practical applications of DRL to active flow control.

/pdf/robust-active-flow-control-over-a-range-of-reynolds-numbers-34ujncfwa7.pdf

Robust active flow control over a range of Reynolds numbers using an artificial neural network trained through deep reinforcement learning

This paper focuses on the active flow control of a computational fluid dynamics simulation over a range of Reynolds numbers using deep reinforcement learning (DRL). More precisely, the proximal policy optimization (PPO) method is used to control the mass flow rate of four synthetic jets symmetrically located on the upper and lower sides of a cylinder immersed in a two-dimensional flow domain. The learning environment supports four flow configurations with Reynolds numbers 100, 200, 300 and 400, respectively. A new smoothing interpolation function is proposed to help the PPO algorithm to learn to set continuous actions, which is of great importance to effectively suppress problematic jumps in lift and allow a better convergence for the training process. It is shown that the DRL controller is able to significantly reduce the lift and drag fluctuations and to actively reduce the drag by approximately 5.7%, 21.6%, 32.7%, and 38.7%, at $Re$=100, 200, 300, and 400 respectively. More importantly, it can also effectively reduce drag for any previously unseen value of the Reynolds number between 60 and 400. This highlights the generalization ability of deep neural networks and is an important milestone to active flow control.

/pdf/robust-active-flow-control-over-a-range-of-reynolds-numbers-3ijf7ojb1v.pdf

Robust active flow control over a range of Reynolds numbers using an artificial neural network trained through deep reinforcement learning.

Geometrical uncertainty in turbomachinery: Tip gap and fillet radius

Dynamic stall represents a challenge in a number of engineering applications including rotorcraft, maneuvering aircraft, gust encounters, and wind turbines. Delay of the onset of dynamic stall and ...

Exploration of High-Frequency Control of Dynamic Stall Using Large-Eddy Simulations

Lift improvements using duty-cycled plasma actuation at low Reynolds numbers

Efforts to reduce blade count and avoid boundary layer separation have led to lowpressure turbine airfoils with significant increases in loading as well as front-loaded pressure distributions. These features have been independently shown to increase losses within the secondary flow field at the end wall. Compound angle blowing from discrete jets on the blade suction surface near the end wall has been shown to be effective in reducing these increased losses and enabling the efficient use of highly loaded blade designs. In this study, experiments are performed on the front loaded L2F low-pressure turbine airfoil in a linear cascade. The required mass flow is reduced by decreasing the hole count from previous configurations and from the introduction of unsteady blowing. The effects of pulsing frequency and duty cycle are investigated using phase-locked stereo particle image velocimetry to demonstrate the large scale movement and hysteresis behavior of the passage vortex interacting with the pulsed jets. Total pressure loss contours at the cascade outlet demonstrate that the efficiency benefit is maintained with the use of unsteady forcing. [DOI: 10.1115/1.4026127]

Parametric Optimization of Unsteady End Wall Blowing on a Highly Loaded Low-Pressure Turbine

Recent applications of active flow control to the low-pressure turbine section of a gas turbine are reviewed and evaluated. Unique aspects of the low-pressure turbine flow environment such as frees...

Active Flow Control for Low-Pressure Turbines

The physics of control by pulsed blowing on a NACA 643-618 natural laminar flow airfoil is studied using hot-film anemometry. Measurements in the uncontrolled separated shear layer indicate that vortex shedding is taking place due to the Kelvin–Helmholtz-type inviscid instability. Steady and pulsed external acoustic excitation is used as well to decouple the frequency content of the perturbation from the vorticity introduced by the jet. Acoustic control introduces either the most unstable frequency or harmonics of carrier frequency in the most unstable range, which amplify exponentially in the separation region yielding a significant delay of the separation location to approximately 55% chord. Experimental data suggest that pulsed jets introduce higher-order harmonics of the low-frequency pulsing as well, amplifying the natural disturbances in the laminar separation. Phase-averaged wavelet analysis is used to study the control physics within a single pulsing period. It is shown that a jet-induced high-fre...

Pulsed Jets Laminar Separation Control Using Instability Exploitation

The present paper focuses on the analysis of a synthetic jet device (with a zero net massflow rate) on a separated boundary layer. Separation has been obtained on a flat plate installed within a converging-diverging test section specifically designed to attain a local velocity distribution typical of a high-lift LPT blade. Both experimental and numerical investigations have been carried out. Unsteady RANS results have been compared with experiments in terms of time-averaged velocity and turbulence intensity distributions. Two different Reynolds number cases have been investigated, namely Re = 200, 000 and Re = 70, 000, which characterize low-pressure turbine operating conditions during take-off/landing and cruise. A range of synthetic jet aerodynamic parameters (Strouhal number and blowing ratio) has been tested in order to analyze the features of control — separated boundary layer interaction for the aforementioned Reynolds numbers.

Turbine blade boundary layer separation suppression via synthetic jet: An experimental and numerical study

A large-amplitude low-frequency oscillation has been reported in the literature for some airfoils in a small range of angles of attack before stall. In the current study, the same low-frequency oscillations are shown to be triggered by two methods of active flow control at poststall angles. Active control is applied in the form of external acoustic excitation or a row of discrete jets that are operated in a pulsed or steady mode. Conditionally averaged particle image velocimetry and time-resolved surface pressure measurements are used to evaluate the spatial and temporal characteristics of the oscillation. By systematically increasing forcing amplitude, each flow control technique is shown to first excite low-frequency oscillation up to a maximum lift increase, after which subsequent increases in forcing amplitude attenuate and ultimately eliminate low-frequency oscillation altogether. At maximum oscillation amplitude, the behavior is highly periodic in the case of acoustic forcing and intermittent in the...

Chiara Bernardini

Papers

Parametric Optimization of Unsteady End Wall Blowing on a Highly Loaded Low-Pressure Turbine

Active Flow Control for Low-Pressure Turbines

Pulsed Jets Laminar Separation Control Using Instability Exploitation

Turbine blade boundary layer separation suppression via synthetic jet: An experimental and numerical study

Large Low-Frequency Oscillations Initiated by Flow Control on a Poststall Airfoil