The aim of this exercise is learn how to prepare, run and analyze results of large-eddy simulations (LES) of external vehicle flows. We go through all steps of the LES process and discuss issues such as spatial and temporal resolution, computational grid, development of the flow, averaging etc.

We start with planning for simulation of flow around simplified ground vehicle. The first step is to define the computational set-up for our LES. Participants will learn how to transfer the experimental set-up or real flow situation on the road to appropriate numerical set-up. Special attention will be put on prescribing appropriate boundary conditions that will correspond to the experimental ones. When the numerical set-up is defined we will learn how to estimate the required near-wall resolution and how to construct a computational grid that satisfies the resolution and numerical requirements of LES. We will discuss different strategies for making computational grids and learn how to concentrate the computational cells in regions where they are most needed. The LES computational grid puts higher requirements on the user than RANS grid. Making a grid that resolves the important near–wall flow and in the same time has reasonable number of nodes is challenging in the flow around complex geometry of vehicle. We will go through different strategies and discuss the issues of grid induced oscillations in the LES flow and how to avoid them. Temporal numerical resolution is also of great importance in LES of flows around vehicles. After defining the spatial resolution the participants will learn how to estimate the time step from points of both numerical aspects and physical processes. Ones we have done the preprocessing for the LES we learn how to set-up the LES solver. Dependent on the needs of the participant in the course we will choose to show how to set-up the LES case in different solvers including commercial packages such as ANSYS, AVL-FIRE, STAR-CCM+ but also open source Open-FOAM. If the participant have their own in-house code we can discuss how to set-up the case in their own code.

Here we will discuss issues such as;

- choice of spatial and temporal numerical scheme
- How to initiate the simulation and how to generate the initial solution?
- How to monitor the progress of the simulation?
- What to save during the simulation?
- How to asses numerical accuracy? How to perform grid refinement study

Several LES simulations of the flow around vehicles has already been performed and prepared for the participant so that they can make exercises on and learn from the simulations about following:

- Did the simulation run properly and does it displays typical LES behaviour?
- What is the actual spatial and temporal resolution in the simulation and what do we need to do to improve the resolution?
- How do we judge if the flow is fully developed in the LES?
- How long do we need to do time averaging and how long do we need to run simulation to obtain the reliable information about the temporal behaviour of the flow (i.e. dominant frequencies)?

The final step in the exercises will be to learn how to post process the result of the LES around ground vehicle. Here we will use a choice instantaneous and time-averaged LES results saved during the run to compute different quantities important in LES. We have prepared MATLAB files that the participants can run and modify in order to solve the different tasks in the exercises. For example we will compute resolved Reynolds stresses and subgrid-scale (SGS) stresses from the instantaneous time histories of the flow. This will be done for two different SGS models used in our simulations, the Standard Smagorinsky model and one dynamics SGS model. Using the results of our exercises we will discuss the applicability of different SGS models in LES for vehicle aerodynamics.

Other sub-tasks that we will do in the exercise contain:

- Computation of energy spectra in different regions of the flow. From results of energy spectra we will judge the resolution but also analyze the time-dependent character of the flow.
- We will compute PDF, cross- and auto-correlation between the signals at different positions in the flow and discuss how these can be used.
- We will compute quantities such as SGS viscosity, SGS turbulence kinetic energy, resolved kinetic energy etc. and analyze their near-wall behaviour.
- We will use the data from the simulation to compute the amount of the turbulent kinetic energy that was resolved and relate it to the theory of LES.

The flow that we will study is generic notch-back vehicle (sedan car) shown in Fig. 1.