Assignment: Computational Fluid Dynamics

Subject: CFD prediction of separated flow over a simplified car model

Problem background: The Ahmed body is a simplified car model widely used by automotive industry to investigate the unsteady separated flow around a car rear-body, carry out detailed data analysis and find out the wake flow development. This assignment study is going to apply a modern computational fluid dynamics (CFD) (ANSYS CFX) approach for solving airflow over two- and three-dimensional Ahmed car model in order to explore the flow behaviour and to understand the underlying flow physics related to the flow problems associated with the vehicle.

Relevant fluid mechanics & CFD knowledge: To be able to solve this problem, the student is expected to be familiar with a number of issues as follows. First, flow separation and underlying physical reasons (viscous, adverse pressure gradient); Second, unsteady wake flow (vortex shedding, Strouhal number, shedding frequency); Third, turbulent flow and CFD modelling of turbulence, its capability and limitations, and Fourth, the effect of moving wall/ground and its modelling.

Task 1  - 2D steady and unsteady RANS studies

To choose an Ahmed car model of known geometry and dimensions with known flow properties and conditions as described in the reference paper ‘Large eddy simulation of flow around the Ahmed body' by C. Hinterberger, M. Garcia-Villalba, and W. Rodi (as seen in figure 1 below). To simplify the problem, you can ignore the four supporting legs used in wind tunnel test. Under this task, you only need to consider a 2D mid-plane shown on top-left. The slope angle has to be sufficient large (i.e. 35o, see test data) in order to have large flow separation along the rear slope.

You can follow similar procedures described in ANSYS CFX tutorial ‘Flow around a blunt body' (note that it uses different dimensions) to sketch the exact dimensions of the Ahmed car model and produce the flow domain around it. When you compare your CFD prediction with test data, pls be aware possible different coordinates origin used, so you may have to ‘shift' your prediction or test data in order to do comparison.

Also be aware if you define reference pressure as pref = 1 atm, this means the CFD predicted pressure is actually the gauge pressure (p-pref), so you just need to divide by the reference dynamic pressure (0.5 ρref Uref2) to get the pressure coefficient Cp for comparison with test data. Cp = (p-pref)/(0.5 ρref Uref2)

For this 2D study, you can use ANSYS student version (max 512k nodes/elements) and generate your mesh in a pure 2D (in this case you have to use Fluent solver) or one to several points in spanwise direction (quasi-3D) to use CFX solver. While mesh quality is important, you still need to carefully control the mesh points not above the student version limit. FET IT team may be able to provide technical support in terms of downloading and installation the software. If any issue occurs, the programme leader is the person to contact for resolving the issues, plus extra H-disk space.

The first step is to use the steady RANS method to predict/show surface pressure and skin friction coefficients, wake velocity profiles, and other flow field features such as viscous layer development along wall surfaces, flow separation, separation bubble size and shape, counter-rotating vortex, etc. Higher marks will be given to those showing a good understanding of and interpreting the results.

To compare results from above with suitable experimental measurements obtained from literature (note that there are publications uploaded on Blackboard for reference. You are encouraged to find more research papers). To adjust some parameters (domain, mesh, turbulence model) within the CFD model so that good agreement is obtained with the results from the steady RANS method in comparison with experimental data or published CFD predictions by other researchers. Higher marks will be given to those who demonstrate proper domain size, proper mesh resolution, and examination of turbulence model sensitivity by at least three popular models.

The second step is to continue with unsteady RANS simulation at fixed time step using same turbulence models. Time-averaged mean results can be compared with steady results and test data to understand the differences and any improvements they may have, and possible underlying physical reasons.

Task 2  - 3D steady RANS study only

The first step is to carry out 3D steady RANS computation by using the parameters obtained from the 2D RANS studies, along both streamwise and wall normal directions. For steady RANS, half an Ahmed car model can be used that can largely reduce computational cost/time. The spanwise domain size can be referenced to other published papers, and suitable grid resolution on the model surface and away from it in flow field. Coursework requirement and marking criteria is similar to those described for 2D steady RANS study.

Although it is not required, some of you may consider applying ‘adaptive meshing' capability, in order to capture dynamic vortex shedding process (no marks given).

The second step (optional)* is to carry out unsteady RANS computation by using the parameters obtained from the steady RANS study (domain, mesh, turbulence model etc.), to compare time-averaged (i.e. mean) results with steady RANS results, to show instantaneous flow field (e.g. contours, velocity vectors, etc.), especially illustrating any vortex shedding patterns.

*For some ambitious students, we encourage you to continue with unsteady 3D RANS or even DES/LES. These studies can be included as appendix not accounting for the 20 page limit. The content will be commented but no mark given.

Extra H-drive can be obtained on the advance requests, due to possible large amount of time history data/files if running transient cases, pls contact

The programme leader will book two extra Labs on March 23, 30 (Wednesdays) for part-time students who want to access full ANSYS version available on campus. This is anticipated that 2D studies have been done by that time using student version.

Materials to be handed in to A-block coursework hub: hard copy of the assignment under one cover.

Deliverables: Specifically, the CW Report should contain, as a minimum, the following sections:

Title/Preface page (not include page limit): assignment title; student name and number; the award programme.

- Abstract: A brief statement to introduce the investigation/work, and to state any research or study findings. Max 300 words or half a page in length.

- A contents page (not necessarily a separate page, if space is limited).

- Introduction of the problem: The Introduction informs the reader what the reasons are for doing the investigation. It sets out the state of knowledge before the investigation commenced and the scope/contents of the study, including literature reviews as well as aims and objectives of the study.

- Problem set-up: details on how to set up problem in CFD context, define boundary conditions, choice correct models, etc.

- Quantity of obtained results: CFD results in lines, contours, streamlines etc.

- Quality of obtained results: assess the accuracy and quality of obtained CFD results, and its competency

- Data analysis, comparisons and discussions: cross comparison of 2D/3D RANS, steady vs unsteady, etc.

- Conclusions: what are findings from this study, and how they are useful for other researchers?

- References: use the Harvard referencing system.

- Report length is max 20 sides of A4, min font size of 11. Note that mark will be reduced if the page limit is exceeded.