Jet mixing optimization using machine learning control

Abstract

We experimentally optimize mixing of a turbulent round jet using machine learning control (MLC) following Li et al (2017). The jet is manipulated with one unsteady minijet blowing in wall-normal direction close to the nozzle exit. The flow is monitored with two hotwire sensors. The first sensor is positioned on the centerline 5 jet diameters downstream of the nozzle exit, i.e. the end of the potential core, while the second is located 3 jet diameters downstream and displaced towards the shear-layer. The mixing performance is monitored with mean velocity at the first sensor. A reduction of this velocity correlates with increased entrainment near the potential core. Machine Learning Control (MLC) is employed to optimize sensor feedback, a general open-loop broadband frequency actuation and combinations of both. MLC has identified the optimal periodic forcing with small duty cycle as the best control policy employing only 400 actuation measurements, each lasting for 5 seconds. This learning rate is comparable if not faster than typical optimization of periodic forcing with two free parameters (frequency and duty cycle). In addition, MLC re-sults indicate that neither new frequencies nor sensor feedback improves mixing further-contrary to many of other turbulence control experiments. The optimality of pure periodic actuation may be attributed to the simple jet flapping mechanism in the minijet plane. The performance of sensor feedback is shown to face a challenge for small duty cycles. The jet mixing results demonstrate the untapped potential of MLC in quickly learning optimal general control policies, even deciding between open-and closed-loop control.

Deep model predictive flow control with limited sensor data and online learning

“…Several machine learning-based approaches have been presented in this context, for instance cluster-based surrogate models (cf. [31], where the drag of an airplane wing was reduced), feedback control laws constructed by genetic programming [45], or reinforcement learning controllers [35]. These approaches are often significantly faster, rendering real-time control feasible.…”

Upstream actuation for bluff-body wake control driven by a genetically inspired optimization

“…These plots represent the relative distance between each snapshot, and it is an indicator of how much the flow evolves from snapshot to snapshot. This visualization was previously used for bluff-body wake flows (Kaiser et al 2014) and flow control studies (Duriez et al 2017;Wu et al 2018). The present plots are coloured using the corresponding drag C d and side force C s coefficients (respectively from left to right).…”

Adaptive separation control of a laminar boundary layer using online dynamic mode decomposition

“…More recently, methods from machine learning have been employed in flow control, in which the control design is framed as a regression problem and solved by genetic algorithms without explicit knowledge of the dynamics (Duriez, Brunton & Noack 2017; Li et al. 2017; Wu 2018). Unfortunately, by design, current model-free approaches exhibit slow learning and adaptation times relative to the fluid dynamic time scales.…”

“…An alternate model-free approach employs downhill simplex optimization to estimate the control parameters for maximizing the time-averaged lift-to-drag ratio of an airfoil (Cattafesta, Tian & Mittal 2009). More recently, methods from machine learning have been employed in flow control, in which the control design is framed as a regression problem and solved by genetic algorithms without explicit knowledge of the dynamics (Duriez, Brunton & Noack 2017;Li et al 2017;Wu 2018). Unfortunately, by design, current model-free approaches exhibit slow learning and adaptation times relative to the fluid dynamic time scales.…”