Khamsin Virtual Racecar challenge 2013 (CFD model racing)

[cdsavage]

Julien, is there a chance you could include some images of the actual mesh used for the simulation? There are some weird artifacts in the mesh in the VTK file.

[N12ck]

this is why we CFD the cars for people to learn more about aero, its not all about the look, its about the data

@cdsavage I will email Julien now

[machin]

OK, so a couple of posts... first my usual number analysis, showing front and rear % and the equivalent "balanced" lift cofficients:-



And the Monza chart showing five results (the remaining are off the chart to left) -I've used the balanced Cl.A figure for this, since this gives a better indication of lap time than the raw "unabalanced" figures.



What's interesting is that the top three cars had completely different aero configurations... low vs medium downforce, similar to last year's real Monza GP, although in our case the fastest car had a sufficiently low enough drag coefficient to come out on top!

[N12ck]

great analysis as always Richard!

I am beginning to work on the 2014 based regulations for much more clarity, any suggestions?

[astracrazy]

quiet happy with my results.

I'm still 5th, but in reality in Magy I would of been a lot lower if it weren't for illegal cars. Now these car are legal and resume position above me I've net gained 2 positions.

seems like my sidepod mods worked which i'm happy about as I spent a lot of time on it.

[julien.decharentenay]

Mesh from the VTK file is as per the CFD run - no downsizing that I am aware of. Variante reported the same issue on the test race at the end of last year and I think that he changed it based on the results. This is the primary reason of us giving the results back to you.

in term of mesh size: the base level is set to 0.25m (the mesh size away from the car). The car and wheels mesh size varies between: 0.25/16 = 1.5625cm and 0.25/32 = 0.78125 cm - the switchover between the two sizes is based on the solver (not my control). I agree that the mesher is not perfect, but it is best I can do at the moment.

A local refinement zone would do the trick, but be too computationally expensive. Forcing the mesh size to 0.78125cm on the front wings would increase significantly the setup time.

If you want the 180mph results, can you send me an email so that I can generate the VTK files for it, upload and send a link.

[cdsavage]

If thats the case I'd really strongly suggest increasing the mesh resolution significantly and dropping 1 or even 2 of the runs to make up for it - I think the end result would be far more accurate than having the 3 runs with the current mesh. As it is basically all the slot gaps on the car are being meshed over, including the rear wing which is at the size specified in the rules.

[Blanchimont]

To either save computing time or to have finer meshes, you could only simulate half the car (not for the 3° yaw case, as it's not symmetrical to the car center line). Don't know if this is already done, but it's possibility to get better results.

From the minimum mesh size of 0,78125cm we can learn that gaps/radii/dimensions should not be in the region of smaller than maybe two or three times the minimum mesh size?

[RicME85]

Focus is now shifting to 2014. Im going to try some things out in the final race that would carry over into next years regs.

[machin]

So, what do I recommend you all do next in preparation for the last race?

First off I would mark your team's CFD results from the first two races on the Silverstone Aero Analysis chart (I recommend you use the "areo balanced" lift coefficients in my previous posts for this purpose):-



This will allow you to choose your optimum "base car" to work with for Silverstone -you want to choose whichever configuration gets you closest to the 1:33 (blue) line (or furthest passed it!).

Once you have picked your base car the next thing you want to do is look at how balanced your design was and rectify that first; most teams need to find more front downforce. The best idea is to conduct some industrial espionage(!) and look at the front wing designs of the cars at the front... and copy those!

Once you have done that I would say to look again at the Silverstone aero analysis chart and decide whether you will benefit most from reducing drag or increasing downforce; whichever one gets you closer to (or further passed) that blue line, and then make some small changes to the model. Personally I would steer clear of trying "intricate" details; I would stick to ensuring that your front wing has 3 well designed (i.e. aero-foil shaped) elements, and go from there.

Good luck!

[cdsavage]

another question about the mesh: is there a boundary layer included?

[julien.decharentenay]

No boundary layer included.

[julien.decharentenay]

Chris, I am running your design with a uniform mesh resolution on the car at 0.78cm to assess the impact on mesh size and performance prediction (one case: head-on at 180mph only). I agree that the CFD modelling is not top-of-the-range, but is, in my opinion, suitable to allow for the designs to be compared against each other. Remember that I need to trade-off between the following: accuracy, robustness, resources and automation.

On another aspect: SketchUp does not generate smooth surfaces as it is a faceted model (points -> lines -> surfaces) and do not support curves. The impact of this can be viewed on the pressure distribution at the wheels for example. This could impair the accuracy of the model (for example, boundary layer detachment) - but I think that this is masked by the coarse resolution used currently.

To all: If you think that the CFD modelling is inadequate and that you can do it better, here is a little challenge: for the next round, I will provide one of the racing car design (same data I get) for you to model alongside your own car design (ie 6 runs) using what you deem to the the appropriate level of detail. You have one week to run both while I will be running all the submitted designs (ie around 36 runs) in the same time-frame. At the end, we can compare results. Let me know if anyone is interested so that I can contact the design team I have in mind to get their agreement to share their design.

[julien.decharentenay]

Forgot to say: CFD modelling is done with first accuracy in space, k-omega SST with wall functions.

[astracrazy]

NIck: Any chance of tidying up and updating the standings page? i.e remove teams that haven't taken part and make it clearer? The original page was much better (just a simple table)