Active control of vortex-induced vibration of a circular cylinder using machine learning

Abstract

We demonstrate the use of high-fidelity computational fluid dynamics simulations in machine-learning based active flow control. More specifically, for the first time, we adopt the genetic programming (GP) to select explicit control laws, in a data-driven and unsupervised manner, for the suppression of vortex-induced vibration (VIV) of a circular cylinder in a low-Reynolds-number flow (Re = 100), using blowing/suction at fixed locations. A cost function that balances both VIV suppression and energy consumption for the control is carefully chosen according to the knowledge obtained from pure blowing/suction open-loop controls. By implementing reasonable constraints to VIV amplitude and actuation strength during the GP evolution, the GP-selected best ten control laws all point to suction-type actuation. The best control law suggests that the suction strength should be nonzero when the cylinder is at its equilibrium position and should increase nonlinearly with the cylinder's transverse displacement. Applying this control law suppresses 94.2% of the VIV amplitude and achieves 21.4% better overall performance than the best open-loop controls. Furthermore, it is found that the GP-selected control law is robust, being effective in flows ranging from Re = 100 to 400. On the contrary, although the P-control can achieve similar performance as the GP-selected control at Re = 100, it deteriorates in higher Reynolds number flows. Although for demonstration purpose the chosen control problem is relatively simple, the training experience and insights obtained from this study can shed some light on future GP-based control of more complicated problems.

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

“…15,21 Such techniques mainly include genetic algorithms (GAs) and artificial neural networks (ANNs). While GAs have been extensively used for AFC, [23][24][25][26] ANNs are receiving growing attention recently due to the fast development of artificial intelligence/machine learning that has taken place in recent years. Furthermore, ANNs have been found so far to surpass GAs in terms of the complexity of the tasks learned and their learning speed.…”

Stabilization of the fluidic pinball with gradient-enriched machine learning control

“…Third, in an ideal scenario, the intended range of operating conditions is already included in the cost function. For instance, a control law may be evaluated at different operating conditions or in a slow transient between them (Asai et al 2019;Ren, Wang & Tang 2019). This will, however, dramatically increase the testing time.…”

Hybrid analysis and modeling, eclecticism, and multifidelity computing toward digital twin revolution

“…For example, GP was applied to control the recirculation area of a backward facing step [115] and mitigate the separation and early reattachment of turbulent boundary layer for a sharp‐edge ramp [80]. Moreover, Ren et al [283] adopted GP to select explicit control laws for the suppression of vortex‐induced vibration of a circular cylinder in a low‐Reynolds‐number flow using blowing/suction at fixed locations. All these ideas can be exploited in emphasizing the potential for using flexible data assimilation and automatic control environments to unlock the capabilities of a DT framework.…”