Khamsin Virtual Racecar Challenge 2015

[MadMatt]

Is there a quick way to know your frontal area from a CAD file or .STL? At the moment I am doing it manually but it is a bit annoying.

I can't think of a simple way. What do you need this for? The results from One-Click CFD include total force and Cd.A / Cl.A, which are independent of reference areas.

As I cannot make One-Click CFD work for now (see the pvbatch error I mentioned previously), I am using another software, but what I get are the downforce and drag values, so I need the frontal area. At the moment I am doing it the slow way by drawing more or less precise shapes in the longitudinal plane to get that area but I thought there would be a software that could "print" the surface in the longitudinal direction onto a plane.

EDIT: Found a feature in my CFD software that gives me the frontal area quite quickly, so case closed, but I was wondering if any of these free software would do the same.

[julien.decharentenay]

@MadMatt Sorry I have been busy trying to get onto OCCFD (I like the accronym). As Chris pointed out, he is handling the compliance and I would have had to handle it to him. Best to check for compliance early.

OCCFD update: I think that I have narrowed down the issue and I have to revert to modelling the full car. I have the very strong feeling that the issue is associated with the symmetry plane in the middle and the mesh. I may be able to do a work around at a later date - but I want to provide you guys with something that works.

Frontal area: there are multiple ways to calculate it depending on software capabilities. In OCCFD it is done using ParaView as follows:
- A front view with parallel projection is created with the car and a black box of approximate dimension that is close to the car (the box is not on top of the car, but above the car);
- Two images are generated: one with the car in black and the box in white and one with the car in white and the box in black. The background is set to white.
- The two images are compared based on the number of black pixels, which gives the frontal area ratio. Important: the images are exported in Bitmap to prevent any filtering.

Simple and quite efficient (if you do not have to do it manually). This being considered, the frontal area is not required and do not impact the laptime as the laptime uses the drag area and downforce area (drag/downforce normalised by the velocity induced pressure)...

[julien.decharentenay]

Quick news: I should have an updated version of one-click CFD ready this Sunday evening (UK time). It should be more or less the final one (at least as far as meshing and boundary conditions parameters are concerned). The updates would be bug fixing. I may give the symmetry another try later down the line, if I can get it to work the change-over would be in the June-August period. @MadMatt: this should sort out the issue you mentioned earlier (and we can probably more on the next set of errors).

[CAEdevice]

Hi Julien, what do you mean with"I may give the symmetry another try later down the line"? In the next version we should use the whole model or half car?
If possibile, it would be useful that OCCFD could manage the symbol " before and after the body name. I can do it with a text editor, but opening and modifying a STL file with a text editor may take a long time.

[RicME85]

Use the whole car, he said recently that half car was possibly the problem

[CAEdevice]

Use the whole car, he said recently that half car was possibly the problem

I'm using the whole car, but with the first OCCFD release I still have a "divergence" issue. I think Julien was talking about the next release.

[julien.decharentenay]

Up to now, OCCFD - I like the acronym - was assuming that you would submit the whole car, but would simulate only half the car by applying a symmetry in the middle. In the next release, you will have to submit the whole car and OCCFD will simulate the whole car.

In the next version (and the one prior): you submit a full car geometry.

Later down the track I will revisit to see if it is possible to reduce the computational cost by applying a symmetry again within OCCFD (you would still submit a whole car). The aim of the symmetry is solely to reduce the computational time (or more accurately enable an increase in mesh size).

The next version includes a name sanitizer. So you will be able to submit the car with the character " in the name (or pretty much any character you want). The character will be removed when doing the simulation by OCCFD.

[CAEdevice]

Thank you. To run the whole car is not a problem. The most important thing is that the results of OCCFD are nearly the same of the race solver.

During development stage a "quick check" option with a less refined mesh could be useful.

[MadMatt]

Matteo I think having a "quick check" is a bit difficult because the only way you can run faster simulations is to use a less refined mesh, and then your results can vary a lot and become hard to trust. You cannot really evaluate the models between each other based on relative difference in the results because these can be affected a lot due to the mesh.

Mesh sensitivity is something that needs to be evaluated and that is very important, but there is always a compromise otherwise we would need 10m cells models that will have to be run on a cluster to have time to simulate everybody's car with a valid mesh.

But I agree with the fact that OCCFD has to be as close as possible to the race solver.

Julien, thank you for sorting the pvbatch issue.

[julien.decharentenay]

Just to confirm OCCFD is the actual race solver.. The one potential difference is that the submitted geometry may (or may not) be going through a mesh cleaning step using netfabb basic prior to being submitted through OCCFD.

[julien.decharentenay]

There you go, the new release candidate is available at http://www.khamsinvirtualracecarchallen ... edirects=0.

Similar to last year, the mesh rough parameters are:
- Cells size externally: 0.5m;
- Cells size at the body surface: ~8mm (= 0.5/64)
- Please do create the wings as separate parts with the name "wing" somewhere (wings-front, wings_rear, my_wing) and then the mesh gets refined to ~4mm (= 0.5/128).

Boundary layers are created on the car body - to the exception of exhausts and inlets - with 2 layers and a layer thickness of 0.15 * cell size.

I will update the presentation... soon...

[julien.decharentenay]

During development stage a "quick check" option with a less refined mesh could be useful.

There is an option you might be able to use. The following line is the command line that is started when you right-click and start OCCFD:

Code: Select all

"{app}\mainOneClickCFDVehicle.exe -fullcar -velocity:44.7 -groundLevel:-0.0445 "PATH"

Where PATH is the path to the STL file or the directory in which the STL files are contained.

Code: Select all

-groundLevel:-0.0445

is to be used to set the ground level to -44.5mm. So it assumes that the underfloor of the car body is at z = 0mm...

If you type the command in a command prompt terminal and add a "-dev" option (as per the command line below), then you can run the solution on a coarse mesh. Let me know if this is something that you are planning to use as I may change the parameters.

Code: Select all

"{app}\mainOneClickCFDVehicle.exe -fullcar -velocity:44.7 -groundLevel:-0.0445 -dev "PATH"

Good luck and let me know if it is working better...

[CAEdevice]

Thank you! Yes I'm going to use the "coarse mesh" option, but it has to be not too much "coarse". I think the a 25% of the average time needed to complete the "standard mesh" run would be ok. The results files would be lighter too.

Maybe it would help who doesn't have much computing power.

It would be better (but only if it does not waste too much of your time) to have that "quick check run" selectable from the right click menu.

[MadMatt]

Thank you Julien, the pvbatch error I had before is now gone, I will try to run a full simulation in the following days!

[julien.decharentenay]

It would be better (but only if it does not waste too much of your time) to have that "quick check run" selectable from the right click menu.

Could you test it a bit and confirm that it is working? Once this is done, I will refine the coarse option a bit (so that the mesh is smaller than the regular, but a bit finer than the current one) and will add this from the right-click menu. It will not show up as a regular right-click, but will show when doing Shift + right-click. This is for the next version, but will not work in the current version!!!