Session Q27: Aerodynamics: Control

Adjoint optimization for optimal actuation on low Re airfoil

Flow control strategies have been shown to favorably modify aerodynamic performance at both high and low Reynolds number. We focus on an actuation strategy in which deformations are driven along the surface of the aerodynamic body. This strategy is promising because it leverages lightweight actuators that can operate across a wide range of timescales. We use two-dimensional high-fidelity simulations of a NACA0012 airfoil at a Reynolds number of 1000 to build on earlier work where the surface deformations were in the form of a travelling wave, parameterized by wavespeed and wavenumber. Here, we use adjoint-based optimization to identify the space- and time-varying deformation profile that maximizes mean lift and minimizes mean drag. We will compare the optimal deformation kinematics to the open-loop travelling wave actuation, and identify non-sinusoidal actuation characteristics that are beneficial to aerodynamic performance. Throughout, the optimal kinematics will be explained in terms of modifications induced in the vortex structures and pressure field near and in the wake of the airfoil.    

Structural Control of an Aerodynamically-Adaptive Flexible Wing using Distributed Bleed Actuation.

Variable aerodynamic loads that are regulated by controlled quasi-static and transitory interactions of distributed bleed actuation with the cross flow over a flexible 3-D wing model are used to alter its structural and aeroelastic characteristics in wind tunnel experiments.  The bleed is driven through arrays of ports on the pressure and suction surfaces by the inherent pressure differences in flight and is regulated using active louvers.  The effects of bleed-induced time-dependent changes in the aerodynamic forces and moments are explored using direct load measurements and surface motion analysis coupled with high-speed stereo PIV in a streamwise-normal plane in the near wake.  These measurements enable assessment of the spanwise sectional lift distribution over the wing reveal using concentrations of streamwise vorticity.  The present investigations demonstrate how spanwise-segmented bleed actuation can excite cross stream oscillations of the wing and independently mitigate these oscillations using the measured tip velocity. Aerodynamic/structural constitutive, multiscale analysis coupled with time dependent distributions of the aerodynamic loads assesses the induced changes in the  apparent structural properties of the wing.   

On the interaction between a leading edge vortex and Active Flow Control on a swept wing

Swept back thin wings used on combat aircraft rely on vortices to augment their lift and provide added control to their control surfaces. However, much of the previous research on active flow control (AFC) focused on the maintenance of attached flow or the forced re-attachment of separated shear layers. Redirection and control of the leading edge vortex (LEV) may also lead to significant improvements in aircraft efficiency and maneuverability. In this study, flow interactions between the LEV and sweeping jet actuators located downstream are investigated in a subsonic wind tunnel on a highly configurable swept wing model. Balance measurements and surface tuft visualizations demonstrated that the formation and propagation of the LEV directly impact AFC efficacy downstream across a wide range of geometric configurations. In turn, AFC can also influence the upstream LEV dynamics through global changes in pressure and velocity fields. Detailed understanding of these interaction mechanisms is crucial for effective exploitation of AFC technology on swept wing aircraft. Future aircraft designs should incorporate AFC into the early conceptual stages so that it can optimally interact with both attached and vortex flows.   

Experimental Reinforcement Learning for Control of Aerodynamic Systems

Systems such as unmanned aerial vehicles (UAVs) and wind turbines can be severely damaged by forces resulting from large atmospheric flow disturbances (e.g. gusts). Machine learning (ML) methods present a possible solution for mitigating these potentially disastrous situations.  Informed by non-intrusive flow measurements in real-time, ML may be used to estimate incoming flow conditions and identify appropriate control strategies for these applications. Reinforcement learning (RL) is a type of ML in which an “agent” may be trained through trial-and-error goal-directed search to identify superior control strategies of dynamical system. A large-scale experimental application of RL is presented, in which a system has been trained to control aerodynamic forces resulting from an unsteady flow. The performance of these methods is analyzed in the context of real-world flow control applications.   

Deep reinforcement learning based control of the wake of a bluff body

We experimentally investigate the manipulation of the near wake of a bluff body using pulsed jets located along the periphery of the model’s base and injected through rectangular slits. The mass flow rate, the duty cycle, and the actuation frequency were the forcing parameters. The actuation is determined using a deep reinforcement learning (DRL) based approach. The neural network is trained using 60 pressure taps that populate the model’s base and the lateral surfaces for a set of freestream conditions while the reward is based on the direct measure of the model’s drag provided by a load cell. We investigate a range of agents, and we show that it learns an optimal forcing leading to drag reduction of the order of 9% within tens of episodes, corresponding to about 3 hours of experiments.

We investigate the capability of the network to adapt to changing conditions, such as different freestream speeds and switching from the case of isolated wake to a platoon marching configuration and the trained agent shows a very good capability to keep satisfactory values of drag reduction.

The comparison of the most energetic POD modes of the microphones located across the model’s base measured in the baseline case (no forcing) and the most favorable forcing condition, reveals that the wake would tend to a symmetric state, suppressing the large-scale oscillations associated with the shedding mechanism, as evidenced looking at the spectra of the temporal coefficients of the POD modes.   

Towards Reliable and Sample Efficient Deep Reinforcement Learning for Aerodynamics : a Benchmark Application

Deep Reinforcement Learning (DRL) recently led to new control solutions for dynamic systems across various domains but its application to Computational Fluid Dynamics (CFD) remains a challenge. DRL algorithms usually require a large number of samples, whereas the high computing cost associated to CFD limits the amount of data which can be produced. Therefore, it is crucial to establish (i) which algorithms are both sample efficient and reliable (ii) how control laws generalize from simple to more sophisticated environments on the same problem. This study targets specifically continuous actions algorithms with a replay buffer: DDPG, TD3 and SAC. The algorithms are trained and evaluated on an aerodynamic benchmark which consists in controlling the trajectory of a 2D airfoil. Trainings performed with a low-order model show that optimal control can be achieved with all three algorithms. Besides, SAC is found to be effective and reliable whereas DDPG and TD3 are prone to instabilities and half of the trainings do not converge to a proper solution. Finally, solutions obtained with the low-order model and with CFD on the same problem are compared and transfer of control solutions between the two modeling methods is discussed.   

Adaptive control of the unsteady loading on a high-rise building

The flow around a high-rise building determines its unsteady wind loading. In this work, the flow structures around a benchmark high-rise building are numerically investigated using Large Eddy Simulation (LES). Different from the finite cylinder in the uniform flow, such buildings protrude into the atmospheric boundary layer, significantly affecting the wake structures. The coherent flow structures are given and the switching process between antisymmetric and symmetric vortex shedding modes are analyzed in detail. To attenuate the unsteady loading, an adaptive control strategy is designed. The control is based on an IIR filter, its coefficients being optimised by the LMS algorithm, and an efficient method to overcome the IIR instability is implemented. This adaptive controller is then implemented in numerical simulations where its performance gets assessed.   

Drag reduction in simplified geometries of blunt vehicles by means of base blowing

We perform a study on the turbulent flow around a square-back blunt model, to analyze the effect on both, the near wake and drag coefficient of base blowing with gases of various densities. Different symmetric and asymmetric base and perimetric blowing configurations are tested, showing that the drag coefficient is more efficiently reduced when the rear perimetrical injection takes place at the base bottom. When different blowing-to-freestream density ratios ρ_b/ρ_∞ are investigated, the comparison reveals that injecting a light fluid provides larger drag reductions than blowing fluid of density equal to or greater than that of the free-stream. Complementary simulations have been conducted, which also show the regularization effect of the blowing on the aerodynamic forces and the wake. Finally, high blowing efficiencies have been also found when the injection takes place at the base centre.   

Bi-stable flows past road vehicles and controlling mechanisms

The wake bi-stability behind a notchback Ahmed body geometry was investigated by performing wind tunnel experiments and Large-eddy simulations (LES). In both approaches, the bi-stable nature showed two stable mirrored states characterized by a left or right asymmetry for long periods. The asymmetry of the flow was ascribed to the asymmetric separations and reattachments in the wake. The mechanism responsible for the stochastic switch between two asymmetric mirrored wake states were explored from the perspective of near-wall flow structures. Modal analysis applying Proper Orthogonal Decomposition (POD) was employed for the analysis of the wake dynamics. The bi-stable phenomenon was found sensitive to the rear effective backlight angle, blockage ratios, yaw angles, and other terms resulting in the change of the wake flow. The flow control of the bi-stable wake was achieved by producing near-wake perturbations in the shear layer for suppression of the bi-stability. The symmetrization of the wake was observed with the vortex generator placed on the centre of the roof, producing vortices interacting with separations attached to the rear slanted surface.