Turbomachinery flows are inherently complex and unsteady with a large proportion
of the flow being turbulent. Turbulence is the unsteady, three-dimensional and
apparent random motion of fluids, responsible for the majority of mixing and
heat transfer and increases of drag and aerodynamic noise generation. Despite
its crucial importance in many applications, the complex nature of turbulence
makes its accurate prediction extremely challenging. This is primarily due to
the fact that turbulent flows are characterized by the presence of a large range
of length and time scales of the eddying motions and their multi-scale interactions.

Direct numerical simulation (DNS), i.e. simulations that resolve all these
spatial and temporal scales, are computationally extremely costly and in many
cases not feasible.

A remedy for this problem has so far been to rely on turbulence modeling, and to
date, the only practical solution for affordable simulation of turbulent flows
in an industrial context has been to model the entire range of turbulence
scales. This approach, called Reynolds-Averaged Navier-Stokes (RANS) has been
able to deliver reliable results for many flows. Unfortunately, there are many
situations where even the most sophisticated RANS models do not perform well,
especially in the presence of large-scale unsteadiness as the one produced by
blade row interactions. Capturing the essential flow physics, and doing so
accurately, is central to a reliable assessment of design benefits and key to
exploration of novel designs for more efficient engines.

In the current project, we will use GPU computing to enable `numerical
experiments’, using no turbulence modeling, of complex turbomachinery flows. Our
in-house high-fidelity flow solver will be modified to fully exploit
GPU-assisted architectures to perform DNS of these flows. The data will be used
to answer basic questions regarding the physics and modeling of the flows
present in modern aeroengines.

This project is being offered as a potential project area for the complex
systems simulation doctoral training centre at the University of Southampton,
please see: http://www.icss.soton.ac.uk/ for more details on the doctoral
programme and the full application process.

The studentship is available to UK or EU students only.

If you wish to discuss any details of this particular project informally, please
contact Prof. R D Sandberg, Aerodynamics and Flight Mechanics research group,
Email: sandberg@soton.ac.uk