[KVRC] Khamsin Virtual Racecar Challenge 2016

[LVDH]

- It would not be the same if you do not participate ( too much desire to beat you on the track )

Yes I know. But right now the situation is not ideal. I would really like to compete, believe me. This way at least the challenge could continue and I could at least promote my new online tool. If it gets reliable enough maybe at some point somebody else can run the official simulations on it. Maybe Julien will return at some point.

- What about geometry check and pre processing (and laptime simulation?) ? I think Chris and Machin job help would be very important

Good question. That is why I want the other current staff members to continue. If they do not agree, continuing the challenge will be difficult.

- What about my proposal of limited checks and simulations to allow you to partecipate ?

Maybe the contenders could submit their expected performance with the vehicle entry. Based on this and our own expectations we could then only check the cars that are expected to get points. Possibly it will cause checks after the first run of simulations but this way we should be able to limit the number of checks. Also MantiumWFlow could take care of some simple checks.

[CAEdevice]

A limited (2 or 3) number of checked cars (geometry and simulations results), maybe randomly selected, and a certain level of trust, should allow a single person (Chris? You? Me? Somebody else?) to do everything. This would allow you to compete.

Anyway: my project (maybe based on single events, not on a challange), would continue marginally to yours, so there would be the possibility to compete sometimes (with Chris too if possible).

[CAEdevice]

PS: I also tried to introduce an AI aspect into the competition (based on Torcs), but a bigger staff (2 or 3 people) would be needed to do this... maybe in the future.

[cdsavage]

I would only ask for a thumbs up from the current KVRC staff to continue and Julien to be OK with this.

You certainly have my thumbs up. I'm going to send you a PM with some more thoughts.

[LVDH]

Thanks, two thumbs to go. Lets keep this challenge alive.

[CAEdevice]

I would only ask for a thumbs up from the current KVRC staff to continue and Julien to be OK with this.

You certainly have my thumbs up. I'm going to send you a PM with some more thoughts.

I'd like to be involed too, in case LDVH will go on with his proposal, I could mantain the sponsorship and cooperate if needed.

[LVDH]

Yes, I hoped you would offer to help.
Just keep in mind that participants cannot be too involved. But help will be needed for sure, the sponsorship as well.

[CAEdevice]

Yes, I hoped you would offer to help.
Just keep in mind that participants cannot be too involved. But help will be needed for sure, the sponsorship as well.

In case your project is confirmed, the sponsorship will be guaranteed, since I am sure about the project quality.

I have some concerns about the approach not to involve the participants into the numeric solution. This could be ok for an "introclass" with simplified rules, but I would look for higher profiles (engineering services providers, university departments, aero students, ...) that can give a contribution. Alternatively you would have to compute all the (10-15 or more) simulations, quite hard work I fear (but it depends on your hardware... if it is fine for you, it is fine from my side as well).

[LVDH]

I am not sure I understand your question.
To ensure a stable transition I would not change the numeric properties for this year and keep all the mesh and solver settings as they are. The only thing I would change is the calculation of the CoP. I do not agree with how Julien "corrected" the CFD results. This will affect the CoP by some 10cm.
For next year I would then use improved settings for the mesh, This should then even allow for second order accuracy numerical schemes. This will then change a lot as the cars will not be so super aggressive anymore and have less down force. However the numbers and flows should look more realistic.

[graham.reeds]

How about just the top 3 being scrutineered? If any drop out the top 3 then the new entrants get checked.

Also a tool to automate the check would be useful. The tool could also create a scrutineered file which lists violations and appends a CRC code so participants could run and submit it themselves. However the tool shouldn't take long to run and could be part of any automation chain.

[CAEdevice]

I am not sure I understand your question.

I was a bit worried about the hard work needed to simulate 10-15 cars.

The only thing I would change is the calculation of the CoP. I do not agree with how Julien "corrected" the CFD results. This will affect the CoP by some 10cm.

I agree. Honestly I did not understand exactly how the CoP was computed, but it was quite different from my estimation (simplified too).

[RicME85]

The corrected CoP was just a shift towards the required CoP of a set amount. Machin said it was the effect of 'dialing in' the car in practice

[LVDH]

Yes, the corrected CoP was something that would be something like track side tuning like you said. This should be kept as is.
There was another correction though. On this one I did not agree. Julien noticed that the numbers out of the CFD simulation did not add up 100%. He corrected that as well. While he made a good observation, I do not agree on how he corrected these numerical inaccuracies. In my opinion his correction was not sound and I am not aware of a better process so I think we have to live with certain inaccuracies that just come out of CFD. But this is only a detail that most non-CFD people would not even notice.

[CAEdevice]

LDVH, can I ask you to explain how the CoP (not the corrected CoP that is ok for me too) was computed in OCCFD?

In my simplified version, the vertical position of the drag component was guessed (assuming it dependent on the frontal area distribution), so it was easy to locate the horizontal position of the dowforce resultant.

[LVDH]

I am not sure if I can get all the details right, right now, let me try:
To calculate the lift distribution you need the overall lift vector and the moment vector around the, in KVRC, X-Axis.
The vector product of the two should be zero. However due to numerical inaccuracies this is not the case in the simulations.
Julien scaled one of the two vectors in order to fix this. However I never understood why should it be vector one and not two or a mixture of the two. As this seemed arbitrary to me I never liked it. In the end CFD only iterates itself to a solution and for numerical reasons analytical equations are never satisfied and you have to live with these non-perfect results. Sometimes you can fix them, in this case the fix did not convince me.