Khamsin Virtual Racecar Challenge 2014

[astracrazy]

that's good to hear - it is something we will look into once we get everything locked down. We will also look into overall floor height because i think a bit more lower will also increase downforce

perhaps you could test that too?

we have some research time with the sim used for the races, but theres no point until we know what formula/rule book we will be using.

**

Just a heads up guys as well, the form I done was a bit two open + there is one or two errors with how i set it up so there will be a new one. I will post it here and e-mail everyone.

[CAEdevice]

The current floor height is realistic, expecially considering the 0.5° rake (the bib is near to touch the ground and it touches the ground when the rake is set to 0,75°)
If we would use a more realistic rake (1°) we should increase and not decrease the floor height measured on the rear wheels plane.

[cdsavage]

While adding rake is a good idea, I think we need to put things into perspective when talking about real-world accuracy of the absolute downforce/drag numbers. I found that small changes to the cell count and distribution had a pretty significant impact on downforce and COP. While this year was a big improvement over 2013, we had around 3million cells per simulation, and I dont think real-world accuracy on geometry as complex as an entire F1 car can really be achieved without increasing the cell count to a level that would be impractical for KVRC. Instead of a significantly higher cell count, I think we should settle on a level that makes sense in terms of sim time and cost, and introduce an adjustment into the laptime calculation if we can find some benchmark to compare against.

[CAEdevice]

While adding rake is a good idea, I think we need to put things into perspective when talking about real-world accuracy of the absolute downforce/drag numbers. I found that small changes to the cell count and distribution had a pretty significant impact on downforce and COP. While this year was a big improvement over 2013, we had around 3million cells per simulation, and I dont think real-world accuracy on geometry as complex as an entire F1 car can really be achieved without increasing the cell count to a level that would be impractical for KVRC. Instead of a significantly higher cell count, I think we should settle on a level that makes sense in terms of sim time and cost, and introduce an adjustment into the laptime calculation if we can find some benchmark to compare against.

I agree. This year I didn't use OpenFoam as solutor (and this was a big limit for my team), but I was wondering if some someone could provide me some data about the hardware and solution time of the "official" CFD model.

[machin]

A few quick comments based on some of the suggestions over the last few pages:-

I found myself spending the majority of my time from round 3 onwards testing on different downforce levels to see where the best laptime was.

Yeah, I can see how this might be a little annoying....

Some ideas:-

Using just high downforce tracks as you suggest might be one idea...

Or we could organise the races so that they start off as low downforce tracks and then build up to the highest downforce track... this way from each race to the next you're simply trying to improve your downforce a bit whilst trying to maintain efficiency. ???

Another option is that we could generate an aero map for each car, i.e. a bit like the chart below, except the "KVRC best limit curve" gets swapped for the best limit of each team's own cars. Prior to each race the competitor declares what position on the curve he wants to play for that event; he doesn't actually have to submit a model that corresponds to that point. The aim of the competitor is then to submit cars which improve their aero-map, not to aim for a very specific set-up for each race. This also means you don't have to submit a car for a race if you don't want to: you could just use a set-up from your existing aero map.



???

If we're introducing fudge factors into the calculation.....

Personally I'm not keen on fudging the results of my lap time calculator if we're using real tracks, since that would basically mean the validity of my program is put into question. What I could do is make up some fictitious tracks which have less reliance on low drag, i.e. which promote high downforce, or we could use actual British Hillclimb courses which are generally very tight and twisty and would promote high downforce set-ups?

It would be great to have a simplified tyre physical model next year...

Virtual Stopwatch already includes a simple tyre model as explained on my website, excerpt below:-

Once the vertical loads on the tyre contact patches are arrived at using the above steps the lateral cornering forces are generated by applying the following equation:- ((A.x^2)+(B.x)+C)*((D.w^2)+(E.w)+F)*G


Where
•x is the vertical load on the tyre.
•A, B and C are unique factors for each chassis, derived from on-track data
•w is the tyre width
•D, E and F are unique factors for each tyre compound, derived from on-track data
•G is a factor relating to the tarmac surface.

It doesn't take into account temperature or wear throughout a lap, but I've found this is unnecessary for the level of accuracy required for amateur to semi-professional levels of motorsport.

[CAEdevice]

Hi Machin, all your considerations are very interesting, thanks!

About the tyres: I read on your website about the tyre model, but (my fault) my thought was not clear. I was actally referring to a wear model, used to simulate more than a lap (possibly an entire race, also considering fuel consumption). The aim would be to have a bigger number of parameters influenced by the CFD performance.

[astracrazy]

sorry guys here is the new poll
https://docs.google.com/forms/d/1ESp9y4 ... =send_form

the old one will not count - my fault.

[julien.decharentenay]

Rake is probably something that could be easily added as an option. But at the end a multi-race competition will be about tailoring the car for the race, which I think is pretty much how it would happen in real-life.

This year I didn't use OpenFoam as solutor (and this was a big limit for my team), but I was wondering if some someone could provide me some data about the hardware and solution time of the "official" CFD model.

Not exactly sure if the "official" CFD model refers to the KVRC simulation. If it does, the hardware is:
- Amazon AWS EC2 m3.large instance - see http://aws.amazon.com/ec2/instance-types/;
- Simulation time varies between approx 5 to 9 hours (from memory) with the majority at either 6 or 7 hours.

[CAEdevice]

Thanks Julien.

An update about the "rake" test:

Rake 0,5° >>> DF +10%
Rake 1,0° >>> DF +15%

The drag value is more or less the same.
Consider that the absolute position of the wings flaps has not been changed with the rotation of the car body and floor.

[CAEdevice]

Not exactly sure if the "official" CFD model refers to the KVRC simulation. If it does, the hardware is:
- Amazon AWS EC2 m3.large instance - see http://aws.amazon.com/ec2/instance-types/;
- Simulation time varies between approx 5 to 9 hours (from memory) with the majority at either 6 or 7 hours.

Yes, I was refering to th KVRC simulation.
"Amazon AWS EC2 m3.large instance" means 2x XEon 5-2670 (20M Cache, 2.60 GHz, 8.00 GT/s Intel® QPI) + 7.5 Gb RAM.
My workstation should be a bit faster (4 vCPU, 64 Gb RAM) but the total required time would be not less than 4 hours.
I think I'll ask you help when I'll configuring it fot Khamsin : )

[machin]

....also considering fuel consumption. The aim would be to have a bigger number of parameters influenced by the CFD performance.

I've been meaning to add fuel consumption data to Virtual Stopwatch for some time, and finally I've got around to it. The examples below show that where aerodynamic efficiency is concerned you can either use the efficiency to increase speed, or to reduce fuel consumption (or do a bit of both), i.e. you can't have your cake AND eat it:-

In the first example, we have two cars, one with "Normal" levels of drag, and the other identical, except it has 10% less drag. You can see that whilst the low drag car obtains higher speeds down the straights, it doesn't use any less fuel (it does save a little amount of fuel during the cornering; about 0.8kg of fuel over an hour; insignificant compared to the total weight of the car/driver/fuel).



In the second example, below, we have the same two cars from the first example, except that this time, the low drag car is short-shifting by 1000rpm to reduce power and hence fuel consumption, however, now his lap time is pegged back to the same time as the higher drag (but higher power and higher fuel consumption) car:-



So from these examples you can see that it is unnecessary to add fuel consumption into the KVRC challenge since it is fairly clear that all competitors would rather use the aerodynamic efficiency to reduce the time taken to cover the allotted course (whether that course be one lap, or multiple laps), since that strategy will allow you to obtain the fastest possible time.

[CAEdevice]

Thank you very much for this analysis, it's clear and complete. I don't want to make you invest too much time, but I was wondering which would be the effect on laptime difference between the two cars if a different engine model were used (fuel flow limit set to 28 g/s plus the electric power for the allowed time, or redistributed to match the total electric energy used during a lap). If your model already takes into account that just don't consider my question.

Another question: it's clear that inertial forces have the most important influence on fuel consumption, but what would happen in a track with little longitudinal inertial forces like Indianapolis?

Thank you very much once again for your patience

[astracrazy]

Thanks to everyone who has filled out the survey - becoming very helpful

Remember if you filled out the first survey you answers will not count, this is a new survey as there was a problem with the first one. So please remember to re-submit - I think there is a few who filled out the first one but haven't filled out the second.

https://docs.google.com/forms/d/1ESp9y4 ... art=invite

[CAEdevice]

Hi astracracy, I filled out both the forms: can you confirm that you received my second one?

[astracrazy]

yes we have