[MVRC] Mantium Virtual Racecar Challenge 2017

[Alonso Fan]

ohh my, f1technical hasn't been sending me emails lately and I've just come to check to find I've missed a whole lot...

whoops

great to hear that the new software is sorted, I was waiting for it to test my new car.

welcome to the new members! i'll probably be very easy to beat!

looking forward to this season with some proper CAD software and all my CFD set up

[Alonso Fan]

here goes...









the car is currently to the 2016 regulations as I finished it a couple of months ago, but will be tweaked for the first race. i'm still not an expert in SolidWorks as you can tell. the nose needs a clean up!

[CAEdevice]

Simply wonderful!

[CAEdevice]

Has someone compared WFlow2016 vs MFlow2017 results? I see 30% differences in DF and 10% in DR. I think that my first 2017 race will be a guess... I will run just 2 or 3 simulations. Anyway, I am happy with that, it looks a big step towards realism.

[variante]

Looks pretty cool indeed, Alonso Fan!

Has someone compared WFlow2016 vs MFlow2017 results? I see 30% differences in DF and 10% in DR. I think that my first 2017 race will be a guess... I will run just 2 or 3 simulations. Anyway, I am happy with that, it looks a big step towards realism.

20% less DF and 15% less drag this year. High efficiency car. Standard settings for both WFlow versions (huge differences in iterations number and, i think, numerical schemes). Didn't simulate exhaust in both cases. Radiators left 100mm thick.

[Alonso Fan]

Thanks guys, glad you like it

[LVDH]

20% less DF and 15% less drag this year. High efficiency car. Standard settings for both WFlow versions (huge differences in iterations number and, i think, numerical schemes). Didn't simulate exhaust in both cases. Radiators left 100mm thick.

Keep in mind that a thicker radiator means more resistance.

[LVDH]

Hi, I have doubts about the pressure (?) resultants on the engine inlet and exhaust.

faceSource faceSource_exhaust_dx output:
weightedAreaAverage(exhaust_dx) for p = 241.23534

>>>Is "p" a pressure or a resultant? Te rulebook uses the words "integral of pressure" [Pa * m^2]

faceSource faceSource_engine_intake output:
weightedAreaAverage(engine_intake) for p = 10.499427

>>>Is "p" a pressure or a resultant? Te rulebook uses the words "integral of pressure" [Pa * m^2]

faceSource mSurf_cooling_mid_dx output:
areaNormalIntegrate(cooling_mid_dx) for U = (-1.4616098 0 0)

>>> OK: this is near to the reuqired flow, as expected (-1.50)

The rest of the results are not too far from the numbers of the old MVRC software (I have the suspect that this new scheme is much more realistic about flow separation).

I suggest to use as much RAM as possibile (for both hardware and virtual hardware): WFlow required about 8-12gb of RAM, MFlow 2017 is now using (the simulation is about half completed) 18-20gb of RAM.

I am not sure why it needs so much more memory. The mesh sizes should be similar.
About the pressure: There is possibly a mistake here. I am sure that the post-processor reports Pa*m^2. But I am not sure if that is what OF writes into the log files. I will have to check Monday morning. For now you should not worry too much. Most of the times the cars were on edge of hitting the allowed limits. So try to make sure that you have a positive number on the inlet and a negative on on the exhaust outlets.

[variante]

20% less DF and 15% less drag this year. High efficiency car. Standard settings for both WFlow versions (huge differences in iterations number and, i think, numerical schemes). Didn't simulate exhaust in both cases. Radiators left 100mm thick.

Keep in mind that a thicker radiator means more resistance.

That's why I mentioned it
I'll test the same car with thinner radiators and see if drag decreases.

[rjsa]

I said exhaust, but it is intake that is crazy here. Very crazy. And my cases do not seem to be converging. I'l keep running variations though, see if I can figure it out.
I got:


faceSource_engine_in_R_000
185.289 Pa

faceSource_exhaust_L_000
67.6322 Pa

About perfornmance, running on about 13.5gb physical, 4gb swab during the meshing phase. A full run is taking 12h against 5 on the previous setup

[CAEdevice]

For the moderators: would it be possible to pin this thread instead of the 2016 threads (KVRC and MVRC)?

[CAEdevice]

I am not sure why it needs so much more memory. The mesh sizes should be similar.
About the pressure: There is possibly a mistake here. I am sure that the post-processor reports Pa*m^2. But I am not sure if that is what OF writes into the log files. I will have to check Monday morning. For now you should not worry too much. Most of the times the cars were on edge of hitting the allowed limits. So try to make sure that you have a positive number on the inlet and a negative on on the exhaust outlets.

The total solution time is not that bad, about 30% more than the time needed in 2016.

About the inlet/exhuasts: I confirm that the data should be "pressure" [Pa]. To compare it with the data included in the rulebook, I think you shoul dmultiply it for the inlet/outlet areas (whole car), in order to obtain [Pa]*[m^2].

It would be great to have a automated check in the HTML report.

Generally speaking, the new software (mesh + numerical schemes) is great and the results muco more realistic.
Vortices are much better simulated and so flow separation.
Thanks again to LDVH for it (even if I will have to re-design the car after the first race).

[rjsa]

The total solution time is not that bad, about 30% more than the time needed in 2016.

Did you try the fast or the full solution? Because the full solution both here at home (4 core OSX laptop, limited to 16gb ram) and at a c4.4xlarge AWS VM (16 core high performance 30gb ram) the time is up north of 100%

And one more question, if you could be so nice
I remember you mentioning non converging models at some time last year, care to tell what was happening? How you fixed it?

[RicME85]

Just had a play around with the CAESES geometry editor for the Intro class and am impressed. If that doesn't get more people participating I don't know what will.

[LVDH]

And one more question, if you could be so nice
I remember you mentioning non converging models at some time last year, care to tell what was happening? How you fixed it?

Last year I was lucky, as I only had to run the cars in question for more iterations. Usually I then also ran the cars close by on the grid for more iterations as well.
Your car is not behaving very well as I have seen now. One thing you can do is have a good look at the flow visualizations and see where flow is separating. These areas should be fixed as they are not good for performance and for actually physical reasons not good for a steady state CFD solver. Then on your car the stl tessellation is not very nice. Maybe you can post which CAD tool you are using and others who are potentially using the same one can help with better settings to export better stl files.
And then it would be nice if you could send me your case to the usual location as I can use it to improve the CFD settings to make them more robust.

Just had a play around with the CAESES geometry editor for the Intro class and am impressed. If that doesn't get more people participating I don't know what will.

The guys from Friendship Systems will be glad to hear that. There will be improvements during the season. I hope a more complicated rear diffuser will come soon. Also we imported the wrong car from CAEdevice, so now the suspension covers are missing. This is at least an easy fix. Also you might have noticed that the app sometimes crashes. Usually just loading again help. This will also get fixed through time.