Drag reduction of a car model by linear genetic programming control

Abstract

We investigate open-and closed-loop active control for aerodynamic drag reduction of a car model. Turbulent flow around a blunt-edged Ahmed body is examined at Re H ≈ 3 × 10 5 based on body height. The actuation is performed with pulsed jets at all trailing edges combined with a Coanda deflection surface. The flow is monitored with pressure sensors distributed at the rear side. We apply a model-free control strategy building on Dracopoulos & Kent (1997) and Gautier et al. (2015). The optimized control laws comprise periodic forcing, multi-frequency forcing and sensor-based feedback including also time-history information feedback and combination thereof. Key enabler is linear genetic programming as simple and efficient framework for multiple inputs (actuators) and multiple outputs (sensors). The proposed linear genetic programming control can select the best open-or closed-loop control in an unsupervised manner. Approximately 33% base pressure recovery associated with 22% drag reduction is achieved in all considered classes of control laws. Intriguingly, the feedback actuation emulates periodic high-frequency forcing by selecting one pressure sensor in the optimal control law. Our control strategy is, in principle, applicable to all multiple actuators and sensors experiments.

Model identification of reduced order fluid dynamics systems using deep learning

“…In the second numerical example, a flow past a cylinder was simulated. The problem domain is nondimensional 50 units in length and nondimensional 10 units in width, and it possesses a cylinder of radius nondimensional 3 units positioned over the point (5,5). Figure 6 shows the computational domain, with a mesh of 3213 nodes.…”

“…Thus, even with moderate complexity problems, the computation cost can still be prohibitive. A number of model identification methods have been presented to improve the capabilities of the mathematical models such as state calibration with 4D Var, 1 neural networks, 2 dynamics with radial basis functions, 3 genetic programming, 2,4,5 and the traditional Galerkin method. 1,6 Reduced order modelling, one of model identification methods, has proven to be a powerful tool to reduce the complexity and large dimensional size of the full discretised dynamical system.…”

Jet mixing optimization using machine learning control

“…Thus, the number of possible duty cycles N DC for a given f is N DC = N sp − 1 = F RT / f − 1, which increases with N sp and decreases with f . This process is similar to that used in Li et al (2017). Figure 2 displays the permitted frequencies and duty cycles and shows the manually selected frequencies which allow for a locally maximum number of duty cycles.…”

“…The larger k ∈ (−1, 1), the smaller the duty cycle. Following Li et al (2017), a much more general multifrequency forcing is considered, generated here by 9 harmonic functions h i = sin(2π f i t), i = 1, . .…”

“…

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.

…”