1. Direct numerical simulations (DNS), where all the turbulence scales are resolved. Even for moderate Reynolds numbers, however, the computational costs required to run a DNS simulation exceed the potential of current computing facilities. DNS should be regarded as a research tool, not a brute-force solution.

2. Reynolds-averaged Navier-Stokes equations (RANS). In this case, the mean flow field is resolved, while the effect of turbulence is modeled by Reynolds stresses. RANS is currently the industry standard implemented in most CFD packages. It performs well for simple stationary flows, but the results must be validated carefully for nonstationary complex flows.

3. Large-eddy simulations (LES). Within this approach, large flow scales are resolved and sub-grid scales modeled. LES can be regarded as a compromise between DNS and RANS and becomes a very promising approach when solving complex flows.

The main goal of the PhD project is to learn the theory of LES, run LES simulations of transient internal flow over a bluff body and compare to results obtained with RANS simulations and laminar models (will be provided).

Ferziger, J. H. (1998), 'Direct and Large Eddy Simulation of Turbulence'.

Ferziger, J. H. & Peric, M. (2001), Computational Methods for Fluid Dynamics, Springer.

Lesieur, M. (2008), Turbulence in Fluids, Springer.

Sagautíguez-Ferran, P. (2004), Turbulence Direct Numerical Simulation and Large-eddy Simulation, John Wiley & Sons, Ltd., chapter 9, pp. 269--299.