Machine Learning Optimisation for Drag Reduction in a Turbulent Boundary Layer

Lead Research Organisation: Imperial College London
Department Name: Dept of Aeronautics

Abstract

The need to reduce the skin-friction drag of aerodynamic vehicles is of paramount importance. Just a 3% reduction in the skin-friction of a long-range commercial aircraft would save £1.2m in jet fuel per year per aircraft and prevent the annual release of 3,000 tonnes of carbon dioxide. However, achieving net skin-friction drag reduction of wall-turbulent flows is notoriously difficult, even at low Reynolds numbers. Despite decades of research, frustrating performance penalties have prevented the development of any functional and economical system for full-scale applications. Any practical turbulent skin friction reduction control strategy must be able to rapidly and autonomously optimise the aerodynamic surface with minimal power input, have far-reaching effects downstream of control, and must be reactive to changes in flow speed. Enabling such a strategy would mark a new era in turbulence control and is a major objective of this PhD project.
During this project, a new type of machine learning paradigm will be developed, which has the ability to optimise different control inputs rapidly, autonomously, adaptively and simultaneously with only a few numerical simulations. This new capability will be exploited to minimise the turbulent surface friction for a flat plate with minimal power input. The propose control strategy is based on wall blowing, which is a very effective method to reduce shear stress and skin-friction drag (up to 70% drag-reduction have been reported in the literature). However, the energy expenditure can be high, leading to net energy saving as low as 5%. The proposed approach will open up the opportunity to rapidly optimise all existing types of actuation based control strategies as well as discover new types of actuation techniques and drag reduction mechanisms to generate significant levels of turbulent skin friction reduction.
This ambitious goal will be achieved by developing and implementing a novel machine learning paradigm based on Bayesian optimisation, a global optimisation technique that only requires few simulations to be performed, and is yet to be exploited in the field of fluid dynamics.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512540/1 01/10/2017 31/03/2022
2092941 Studentship EP/R512540/1 01/10/2017 30/09/2021 Omar Mahfoze
 
Description We performed Bayesian optimisation studies (to optimise some parameters of the blowing solution such as intensity, streamwise extent, intermittency, etc.) and we found that a net-power saving on the order of a few percent is possible using a low-amplitude wall-normal blowing control strategy at low to moderate Reynolds numbers for a zero-pressure gradient turbulent boundary layer.
Exploitation Route Method is available to the scientific community
Sectors Aerospace, Defence and Marine,Energy

 
Description [EnAble]: Developing and Exploiting Intelligent Approaches for Turbulent Drag Reduction
Amount £498,547 (GBP)
Funding ID EP/T021144/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Public
Country United Kingdom
Start 10/2020 
End 09/2023
 
Description Airbus 
Organisation Airbus Group
Country France 
Sector Academic/University 
PI Contribution City University London: collaboration with AIRBUS to study the characterisation of swept wings in transitional and turbulent regimes and the introduction of large scale vortex generators to control separation at high loading.
Collaborator Contribution City University London: collaboration with AIRBUS to study the characterisation of swept wings in transitional and turbulent regimes and the introduction of large scale vortex generators to control separation at high loading.
Impact See list of publications
Start Year 2019
 
Title Incompact3d 
Description Incompact3d is a powerful high-order flow solver for academic research. Dedicated to Direct and Large Eddy Simulations (DNS/LES), it can combine the versatility of industrial codes with the accuracy of spectral codes. It scales with up to one million cores. The incompressible Navier-Stokes equations are discretized with finite-difference sixth-order schemes on a Cartesian mesh. Explicit or semi-implicit temporal schemes can be used for the time advancement depending on the flow configuration. To treat the incompressibility condition, a fractional step method requires to solve a Poisson equation. This equation is fully solved in spectral space via the use of relevant 3D Fast Fourier transforms(FFTs), allowing any kind of boundary conditions for the velocity field in each spatial direction. Using the concept of the modified wavenumber, the divergence free condition is ensured up to machine accuracy. The pressure field is staggered from the velocity field by half a mesh to avoid spurious oscillations. The modelling of a solid body inside the computational domain is performed with a customised Immersed Boundary Method. It is based on a direct forcing to ensure a no-slip boundary condition at the wall of the solid body while imposing non-zero velocities inside the solid body to avoid discontinuities on the velocity field. This customised IBM, fully compatible with the 2D domain decomposition and with a possible mesh refinement at the wall, is based on a 1D expansion of the velocity field from fluid regions into solid regions using Lagrange polynomials. To reach realistic Reynolds numbers, an implicit LES strategy can be implemented to solve the Navier-Stokes equations without any extra explicit modelling. In order to mimic a subgrid-scale model, artificial dissipation can be added via the viscous term thanks to the artificial dissipative features of the high-order compact schemes. 
Type Of Technology Software 
Year Produced 2019 
Open Source License? Yes  
Impact see list of publications 
URL http://www.incompact3d.com